The Kijulo action guide

From feature name to first useful result

Explore what Kijulo can do, where each capability lives, and the shortest path to using it. Every use case links back here when a step needs one of these building blocks.

Explore features by domain

DB

Organize

Create structured data

Reusable setup

Model the people, projects, notes, opportunities, or any other things you care about as typed, searchable rows.

Doing it manually

20–45 min

Delegated to an agent

3–8 min

See a prompt example

Help me: Create structured data. Goal: Model the people, projects, notes, opportunities, or any other things you care about as typed, searchable rows. Use the relevant data and tools available in Data → New entity type. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Data → New entity type

How to get started

  1. 1 Name the kind of thing you want to track.
    • Where: Open Data → New entity type.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Name the kind of thing you want to track.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Add the fields, types, validation, and relationships it needs.
    • Where: Continue in Data → New entity type.
    • Configure: Add the fields, types, validation, and relationships it needs.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Create rows manually, import them, or let an agent propose them.
    • Where: Complete this step in Data → New entity type.
    • Run: Create rows manually, import them, or let an agent propose them.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
IN

Organize

Ingest a video, PDF, or document

Repeatable action

Turn YouTube transcripts, PDFs, scans, images, and pasted text into editable markdown, then extract reusable knowledge from it.

Doing it manually

15–60 min

Delegated to an agent

2–8 min

See a prompt example

Help me: Ingest a video, PDF, or document. Goal: Turn YouTube transcripts, PDFs, scans, images, and pasted text into editable markdown, then extract reusable knowledge from it. Use the relevant data and tools available in Smart Ingest. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Smart Ingest

How to get started

  1. 1 Submit a YouTube URL, PDF, or document folder; scanned PDFs can be processed with OCR.
    • Where: Open Smart Ingest.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Submit a YouTube URL, PDF, or document folder; scanned PDFs can be processed with OCR.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Wait for Kijulo to preserve the source and create a normalized markdown artifact.
    • Where: Continue in Smart Ingest.
    • Configure: Wait for Kijulo to preserve the source and create a normalized markdown artifact.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Open the markdown, review it, then extract facts, ideas, and relationships into Entity Types.
    • Where: Complete this step in Smart Ingest.
    • Run: Open the markdown, review it, then extract facts, ideas, and relationships into Entity Types.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
FS

Organize

Work with local files

Reusable setup

Give an agent explicit access to selected files and folders while keeping everything portable on disk.

Doing it manually

10–20 min

Delegated to an agent

2–5 min

See a prompt example

Help me: Work with local files. Goal: Give an agent explicit access to selected files and folders while keeping everything portable on disk. Use the relevant data and tools available in Agent conversation → Context, and Settings → LLM → File access. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent conversation → Context, and Settings → LLM → File access

How to get started

  1. 1 Add folders to the Read Allowlist and, when needed, the Write Allowlist.
    • Where: Open Agent conversation → Context, and Settings → LLM → File access.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Add folders to the Read Allowlist and, when needed, the Write Allowlist.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Start a conversation and select the exact files or folders in scope.
    • Where: Continue in Agent conversation → Context, and Settings → LLM → File access.
    • Configure: Start a conversation and select the exact files or folders in scope.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Review every proposed file edit before applying it.
    • Where: Complete this step in Agent conversation → Context, and Settings → LLM → File access.
    • Run: Review every proposed file edit before applying it.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
AI

Think and decide

Start an agent conversation

Repeatable action

Ask an AI to reason over only the rows and files you choose, with an explicit posture and bounded tools.

Doing it manually

30–90 min

Delegated to an agent

3–15 min

See a prompt example

Help me: Start an agent conversation. Goal: Ask an AI to reason over only the rows and files you choose, with an explicit posture and bounded tools. Use the relevant data and tools available in Agent → New conversation. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent → New conversation

How to get started

  1. 1 Choose a posture that matches how much the agent may do.
    • Where: Open Agent → New conversation.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Choose a posture that matches how much the agent may do.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Select the Entity Types, rows, files, and folders it may use.
    • Where: Continue in Agent → New conversation.
    • Configure: Select the Entity Types, rows, files, and folders it may use.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Write your goal, then review any proposed changes in the Review queue.
    • Where: Complete this step in Agent → New conversation.
    • Run: Write your goal, then review any proposed changes in the Review queue.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
WF

Automate

Create a Workflow

Reusable setup

Turn a useful sequence into a runnable recipe made of prompts, templates, and tool calls.

Doing it manually

30–60 min

Delegated to an agent

5–15 min

See a prompt example

Help me: Create a Workflow. Goal: Turn a useful sequence into a runnable recipe made of prompts, templates, and tool calls. Use the relevant data and tools available in Library → Workflows. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Library → Workflows

How to get started

  1. 1 Choose what triggers the Workflow: rows, files, or a tag.
    • Where: Open Library → Workflows.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Choose what triggers the Workflow: rows, files, or a tag.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Add and order its template, tool-call, and user-prompt steps.
    • Where: Continue in Library → Workflows.
    • Configure: Add and order its template, tool-call, and user-prompt steps.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Set tool and file permissions, test it, then run or schedule it.
    • Where: Complete this step in Library → Workflows.
    • Run: Set tool and file permissions, test it, then run or schedule it.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
FIT

Think and decide

Create a Fit Plan

Reusable setup

Rank rows from one table against candidates in another using transparent, weighted criteria.

Doing it manually

30–60 min

Delegated to an agent

5–10 min

See a prompt example

Help me: Create a Fit Plan. Goal: Rank rows from one table against candidates in another using transparent, weighted criteria. Use the relevant data and tools available in Fit Plans → New Fit Plan. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Fit Plans → New Fit Plan

How to get started

  1. 1 Pick the source table and one or more target tables.
    • Where: Open Fit Plans → New Fit Plan.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Pick the source table and one or more target tables.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Map comparable fields and assign weights or matching rules.
    • Where: Continue in Fit Plans → New Fit Plan.
    • Configure: Map comparable fields and assign weights or matching rules.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Run the plan, inspect the criterion breakdown, and optionally record results through an output binding.
    • Where: Complete this step in Fit Plans → New Fit Plan.
    • Run: Run the plan, inspect the criterion breakdown, and optionally record results through an output binding.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Run the Fit Plan and view the generated matches

PR

Think and decide

Create a Protocol

Reusable setup

Standardize reviews and decisions with reusable, typed questions instead of unstructured prose.

Doing it manually

20–45 min

Delegated to an agent

5–10 min

See a prompt example

Help me: Create a Protocol. Goal: Standardize reviews and decisions with reusable, typed questions instead of unstructured prose. Use the relevant data and tools available in Library → Protocols → Create Protocol. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Library → Protocols → Create Protocol

How to get started

  1. 1 Define the review or decision the Protocol should standardize.
    • Where: Open Library → Protocols → Create Protocol.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Define the review or decision the Protocol should standardize.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Add ordered steps with yes/no, scale, list, matrix, or free-text prompts.
    • Where: Continue in Library → Protocols → Create Protocol.
    • Configure: Add ordered steps with yes/no, scale, list, matrix, or free-text prompts.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Run it for each case and keep the structured answers for comparison.
    • Where: Complete this step in Library → Protocols → Create Protocol.
    • Run: Run it for each case and keep the structured answers for comparison.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
CR

Automate

Schedule recurring work

Reusable setup

Run a Workflow automatically on a calendar or cron schedule with controlled inputs and approval rules.

Doing it manually

10–20 min

Delegated to an agent

2–5 min

See a prompt example

Help me: Schedule recurring work. Goal: Run a Workflow automatically on a calendar or cron schedule with controlled inputs and approval rules. Use the relevant data and tools available in Planning → New Schedule. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Planning → New Schedule

How to get started

  1. 1 Select the Workflow and define its row, query, or folder inputs.
    • Where: Open Planning → New Schedule.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Select the Workflow and define its row, query, or folder inputs.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Choose a frequency, timezone, and optional model override.
    • Where: Continue in Planning → New Schedule.
    • Configure: Choose a frequency, timezone, and optional model override.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Decide whether writes require approval, then monitor runs in Planning and Orchestration.
    • Where: Complete this step in Planning → New Schedule.
    • Run: Decide whether writes require approval, then monitor runs in Planning and Orchestration.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
API

Connect

Connect an external API

Reusable setup

Pull fresh data from an HTTP API on demand or on a schedule and map it into Kijulo rows.

Doing it manually

30–90 min

Delegated to an agent

10–25 min

See a prompt example

Help me: Connect an external API. Goal: Pull fresh data from an HTTP API on demand or on a schedule and map it into Kijulo rows. Use the relevant data and tools available in Library → API Connectors. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Library → API Connectors

How to get started

  1. 1 Choose a target Entity Type and configure the request and credential.
    • Where: Open Library → API Connectors.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Choose a target Entity Type and configure the request and credential.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Use Dry run to inspect the response, then map response fields to columns.
    • Where: Continue in Library → API Connectors.
    • Configure: Use Dry run to inspect the response, then map response fields to columns.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Call it from an agent or schedule it to keep rows current.
    • Where: Complete this step in Library → API Connectors.
    • Run: Call it from an agent or schedule it to keep rows current.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
EM

Connect

Run an outreach Campaign

Repeatable action

Send templated, personalized messages to selected rows and keep outreach tied to structured data.

Doing it manually

45–120 min

Delegated to an agent

10–25 min

See a prompt example

Help me: Run an outreach Campaign. Goal: Send templated, personalized messages to selected rows and keep outreach tied to structured data. Use the relevant data and tools available in Campaigns and Email Templates. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Campaigns and Email Templates

How to get started

  1. 1 Create an Email Template with placeholders from your Entity Type.
    • Where: Open Campaigns and Email Templates.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Create an Email Template with placeholders from your Entity Type.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Choose the target rows or filtered audience for the Campaign.
    • Where: Continue in Campaigns and Email Templates.
    • Configure: Choose the target rows or filtered audience for the Campaign.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Review the messages, send through a connected provider, and track outcomes.
    • Where: Complete this step in Campaigns and Email Templates.
    • Run: Review the messages, send through a connected provider, and track outcomes.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

DOC

Organize

Search and synthesize documents

Repeatable action

Index source files so an agent can retrieve relevant passages and produce grounded summaries.

Doing it manually

30–120 min

Delegated to an agent

5–15 min

See a prompt example

Help me: Search and synthesize documents. Goal: Index source files so an agent can retrieve relevant passages and produce grounded summaries. Use the relevant data and tools available in Files, Smart Ingest, and Agent conversations. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Files, Smart Ingest, and Agent conversations

How to get started

  1. 1 Add source documents inside an allowed folder.
    • Where: Open Files, Smart Ingest, and Agent conversations.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Add source documents inside an allowed folder.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Convert scans or PDFs to markdown when needed and let Kijulo index them.
    • Where: Continue in Files, Smart Ingest, and Agent conversations.
    • Configure: Convert scans or PDFs to markdown when needed and let Kijulo index them.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Start a scoped conversation and ask for a synthesis with links back to sources.
    • Where: Complete this step in Files, Smart Ingest, and Agent conversations.
    • Run: Start a scoped conversation and ask for a synthesis with links back to sources.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
VW

Organize

Build a useful view

Reusable setup

Turn rows into focused tables, boards, filters, groups, and saved views for daily work.

Doing it manually

10–25 min

Delegated to an agent

2–5 min

See a prompt example

Help me: Build a useful view. Goal: Turn rows into focused tables, boards, filters, groups, and saved views for daily work. Use the relevant data and tools available in Open an Entity Type → View switcher. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Open an Entity Type → View switcher

How to get started

  1. 1 Choose table, board, tree, or calendar based on the job.
    • Where: Open Open an Entity Type → View switcher.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Choose table, board, tree, or calendar based on the job.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Filter, sort, group, and show only the columns you need.
    • Where: Continue in Open an Entity Type → View switcher.
    • Configure: Filter, sort, group, and show only the columns you need.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Save the configuration as a named view you can reopen.
    • Where: Complete this step in Open an Entity Type → View switcher.
    • Run: Save the configuration as a named view you can reopen.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
MAP

Organize

Map and compare locations

Reusable setup

Plot rows with geographic fields, select areas, and analyze spatial groups.

Doing it manually

20–45 min

Delegated to an agent

5–10 min

See a prompt example

Help me: Map and compare locations. Goal: Plot rows with geographic fields, select areas, and analyze spatial groups. Use the relevant data and tools available in Spatial. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Spatial

How to get started

  1. 1 Add a geo_point field to an Entity Type and populate its rows.
    • Where: Open Spatial.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Add a geo_point field to an Entity Type and populate its rows.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Open Spatial and select the layer or dataset to display.
    • Where: Continue in Spatial.
    • Configure: Open Spatial and select the layer or dataset to display.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Select areas, save views, or ask an agent to compare the selected rows.
    • Where: Complete this step in Spatial.
    • Run: Select areas, save views, or ask an agent to compare the selected rows.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

CAL

Organize

Plan dates on the Calendar

Reusable setup

See deadlines, visits, events, and recurring work from multiple Entity Types in one calendar.

Doing it manually

15–30 min

Delegated to an agent

3–8 min

See a prompt example

Help me: Plan dates on the Calendar. Goal: See deadlines, visits, events, and recurring work from multiple Entity Types in one calendar. Use the relevant data and tools available in Calendar. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Calendar

How to get started

  1. 1 Add date or datetime fields to the Entity Types you want to plan.
    • Where: Open Calendar.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Add date or datetime fields to the Entity Types you want to plan.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Create a calendar layer for each relevant field.
    • Where: Continue in Calendar.
    • Configure: Create a calendar layer for each relevant field.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Choose a time range and save useful layer combinations as views.
    • Where: Complete this step in Calendar.
    • Run: Choose a time range and save useful layer combinations as views.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
CMD

Automate

Record a proven command or runbook

Reusable setup

Keep working commands and operational procedures so agents can reuse what has already been verified.

Doing it manually

15–30 min

Delegated to an agent

3–8 min

See a prompt example

Help me: Record a proven command or runbook. Goal: Keep working commands and operational procedures so agents can reuse what has already been verified. Use the relevant data and tools available in Command Book. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Command Book

How to get started

  1. 1 Record the command or ordered procedure and the context where it applies.
    • Where: Open Command Book.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Record the command or ordered procedure and the context where it applies.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Verify it against reality and preserve the resulting evidence.
    • Where: Continue in Command Book.
    • Configure: Verify it against reality and preserve the resulting evidence.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Let future agents look it up before guessing or repeating discovery work.
    • Where: Complete this step in Command Book.
    • Run: Let future agents look it up before guessing or repeating discovery work.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

OK

Review and control

Review and approve agent changes

Repeatable action

Keep human control over proposed row, file, schema, and external changes before they take effect.

Doing it manually

10–30 min

Delegated to an agent

2–8 min

See a prompt example

Help me: Review and approve agent changes. Goal: Keep human control over proposed row, file, schema, and external changes before they take effect. Use the relevant data and tools available in Review. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Review

How to get started

  1. 1 Open the relevant Entities, Files, Schema, or Outbox tab.
    • Where: Open Review.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Open the relevant Entities, Files, Schema, or Outbox tab.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Inspect the proposed values, source, confidence, and any conflicts.
    • Where: Continue in Review.
    • Configure: Inspect the proposed values, source, confidence, and any conflicts.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Edit if needed, then approve only the fields you want or reject the proposal.
    • Where: Complete this step in Review.
    • Run: Edit if needed, then approve only the fields you want or reject the proposal.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Read the full documentation →
SYNC

Automate

Call an API on a schedule

Reusable setup

Fetch fresh information automatically and record each result as structured, traceable rows in your database.

Doing it manually

30–90 min

Delegated to an agent

5–15 min

See a prompt example

Help me: Call an API on a schedule. Goal: Fetch fresh information automatically and record each result as structured, traceable rows in your database. Use the relevant data and tools available in Library → API Connectors, then Planning → Schedules. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Library → API Connectors, then Planning → Schedules

How to get started

  1. 1 Configure and dry-run the API request, authentication, pagination, and target table.
    • Where: Open Library → API Connectors, then Planning → Schedules.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Configure and dry-run the API request, authentication, pagination, and target table.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Map response fields to columns and choose a stable key for updates and deduplication.
    • Where: Continue in Library → API Connectors, then Planning → Schedules.
    • Configure: Map response fields to columns and choose a stable key for updates and deduplication.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Attach the connector to a schedule, then review run history and failed records.
    • Where: Complete this step in Library → API Connectors, then Planning → Schedules.
    • Run: Attach the connector to a schedule, then review run history and failed records.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

CLU

Think and decide

Identify clusters in your data

Repeatable action

Discover natural groups, recurring themes, and outliers across rows without classifying every item by hand.

Doing it manually

60–180 min

Delegated to an agent

5–20 min

See a prompt example

Help me: Identify clusters in your data. Goal: Discover natural groups, recurring themes, and outliers across rows without classifying every item by hand. Use the relevant data and tools available in Agent conversation → selected table or view. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent conversation → selected table or view

How to get started

  1. 1 Select the rows and meaningful fields the analysis should consider.
    • Where: Open Agent conversation → selected table or view.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Select the rows and meaningful fields the analysis should consider.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Ask the agent to propose clusters, labels, defining traits, and outliers with evidence.
    • Where: Continue in Agent conversation → selected table or view.
    • Configure: Ask the agent to propose clusters, labels, defining traits, and outliers with evidence.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Review the grouping and save accepted cluster labels or relationships back to the table.
    • Where: Complete this step in Agent conversation → selected table or view.
    • Run: Review the grouping and save accepted cluster labels or relationships back to the table.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

A/B

Think and decide

Compare subsets of your data

Repeatable action

Contrast cohorts, periods, regions, candidates, or any two filtered groups using the same explicit criteria.

Doing it manually

30–90 min

Delegated to an agent

5–15 min

See a prompt example

Help me: Compare subsets of your data. Goal: Contrast cohorts, periods, regions, candidates, or any two filtered groups using the same explicit criteria. Use the relevant data and tools available in Saved Views → Agent conversation. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Saved Views → Agent conversation

How to get started

  1. 1 Create a saved view or filter for each subset you want to compare.
    • Where: Open Saved Views → Agent conversation.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Create a saved view or filter for each subset you want to compare.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Define the measures, qualitative criteria, and time period that must stay consistent.
    • Where: Continue in Saved Views → Agent conversation.
    • Configure: Define the measures, qualitative criteria, and time period that must stay consistent.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Ask the agent for differences, similarities, outliers, and source-linked conclusions.
    • Where: Complete this step in Saved Views → Agent conversation.
    • Run: Ask the agent for differences, similarities, outliers, and source-linked conclusions.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

CSV

Organize

Import Excel or CSV data

Repeatable action

Bring spreadsheet data into typed Kijulo tables while preserving headers, detecting formats, and surfacing invalid rows.

Doing it manually

30–120 min

Delegated to an agent

5–20 min

See a prompt example

Help me: Import Excel or CSV data. Goal: Bring spreadsheet data into typed Kijulo tables while preserving headers, detecting formats, and surfacing invalid rows. Use the relevant data and tools available in Data → Import. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Data → Import

How to get started

  1. 1 Choose the Excel or CSV file and preview its sheets, headers, and detected values.
    • Where: Open Data → Import.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Choose the Excel or CSV file and preview its sheets, headers, and detected values.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Map columns to an existing Entity Type or create a suitable structure for the import.
    • Where: Continue in Data → Import.
    • Configure: Map columns to an existing Entity Type or create a suitable structure for the import.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Validate types and duplicates, import valid rows, then review anything that needs correction.
    • Where: Complete this step in Data → Import.
    • Run: Validate types and duplicates, import valid rows, then review anything that needs correction.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

REL

Organize

Create relationships between data

Repeatable action

Connect people, projects, files, events, ideas, and other rows so context can be reused across tasks.

Doing it manually

20–60 min

Delegated to an agent

3–10 min

See a prompt example

Help me: Create relationships between data. Goal: Connect people, projects, files, events, ideas, and other rows so context can be reused across tasks. Use the relevant data and tools available in Entity Type schema → Relationship field. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Entity Type schema → Relationship field

How to get started

  1. 1 Choose the two Entity Types to connect and name what the relationship means.
    • Where: Open Entity Type schema → Relationship field.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Choose the two Entity Types to connect and name what the relationship means.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Set its direction and cardinality, then add the relationship field to the schema.
    • Where: Continue in Entity Type schema → Relationship field.
    • Configure: Set its direction and cardinality, then add the relationship field to the schema.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Link rows manually or let an agent propose matches for your review.
    • Where: Complete this step in Entity Type schema → Relationship field.
    • Run: Link rows manually or let an agent propose matches for your review.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

WEB

Connect

Research the web into your database

Repeatable action

Let an agent search the web, extract relevant facts, and propose sourced rows instead of leaving findings in a chat.

Doing it manually

60–240 min

Delegated to an agent

10–30 min

See a prompt example

Help me: Research the web into your database. Goal: Let an agent search the web, extract relevant facts, and propose sourced rows instead of leaving findings in a chat. Use the relevant data and tools available in Agent conversation → Web search tool. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent conversation → Web search tool

How to get started

  1. 1 Select the target table and define the fields, scope, freshness, and trusted sources.
    • Where: Open Agent conversation → Web search tool.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Select the target table and define the fields, scope, freshness, and trusted sources.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Ask the agent to search, compare sources, and attach a URL and retrieval date to every claim.
    • Where: Continue in Agent conversation → Web search tool.
    • Configure: Ask the agent to search, compare sources, and attach a URL and retrieval date to every claim.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Review duplicates and confidence, then approve the proposed rows and relationships.
    • Where: Complete this step in Agent conversation → Web search tool.
    • Run: Review duplicates and confidence, then approve the proposed rows and relationships.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

Automate

Run multiple agent conversations at once

Repeatable action

Delegate independent research or processing tasks concurrently and bring their structured results back into one review flow.

Doing it manually

60–180 min

Delegated to an agent

10–30 min

See a prompt example

Help me: Run multiple agent conversations at once. Goal: Delegate independent research or processing tasks concurrently and bring their structured results back into one review flow. Use the relevant data and tools available in Agent → Conversations and Orchestration. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent → Conversations and Orchestration

How to get started

  1. 1 Split the goal into independent scopes with clear inputs, outputs, and permissions.
    • Where: Open Agent → Conversations and Orchestration.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Split the goal into independent scopes with clear inputs, outputs, and permissions.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Start the conversations concurrently and monitor progress, tool calls, and failures.
    • Where: Continue in Agent → Conversations and Orchestration.
    • Configure: Start the conversations concurrently and monitor progress, tool calls, and failures.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Review, reconcile, and approve their proposals before combining the results.
    • Where: Complete this step in Agent → Conversations and Orchestration.
    • Run: Review, reconcile, and approve their proposals before combining the results.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

GO

Automate

Create a focused Agent Launcher

Reusable setup

Save a persona, model, tool preset, skills, and narrow context so repeated conversations start faster and use fewer tokens.

Doing it manually

15–30 min

Delegated to an agent

2–5 min

See a prompt example

Help me: Create a focused Agent Launcher. Goal: Save a persona, model, tool preset, skills, and narrow context so repeated conversations start faster and use fewer tokens. Use the relevant data and tools available in Agent → Launchers. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent → Launchers

How to get started

  1. 1 Name the recurring job and choose the agent persona, model, and instructions.
    • Where: Open Agent → Launchers.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Name the recurring job and choose the agent persona, model, and instructions.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Enable only the tools and skills it needs, and restrict its default data and file context.
    • Where: Continue in Agent → Launchers.
    • Configure: Enable only the tools and skills it needs, and restrict its default data and file context.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Test the Launcher on a representative task, then reuse it whenever that job returns.
    • Where: Complete this step in Agent → Launchers.
    • Run: Test the Launcher on a representative task, then reuse it whenever that job returns.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

TRUST

Review and control

Improve data reliability

Reusable setup

Add ownership and verification signals, then protect trusted files, rows, and tables from accidental changes.

Doing it manually

20–60 min

Delegated to an agent

5–15 min

See a prompt example

Help me: Improve data reliability. Goal: Add ownership and verification signals, then protect trusted files, rows, and tables from accidental changes. Use the relevant data and tools available in Entity fields, file metadata, and lock controls. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Entity fields, file metadata, and lock controls

How to get started

  1. 1 Record authorship and add reviewed and verified fields with clear definitions.
    • Where: Open Entity fields, file metadata, and lock controls.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Record authorship and add reviewed and verified fields with clear definitions.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Review existing records, preserve their sources, and set the appropriate reliability status.
    • Where: Continue in Entity fields, file metadata, and lock controls.
    • Configure: Review existing records, preserve their sources, and set the appropriate reliability status.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Lock trusted files, rows, or tables and require explicit review before future changes.
    • Where: Complete this step in Entity fields, file metadata, and lock controls.
    • Run: Lock trusted files, rows, or tables and require explicit review before future changes.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action

>_

Coding

Share a live terminal with your agent

Repeatable action

Watch the agent’s commands and output as they run, type into the same persistent terminal yourself, and deliberately share your own command results back with the agent.

Doing it manually

10–60 min

Delegated to an agent

2–15 min

See a prompt example

Help me: Share a live terminal with your agent. Goal: Watch the agent’s commands and output as they run, type into the same persistent terminal yourself, and deliberately share your own command results back with the agent. Use the relevant data and tools available in Agent conversation → Terminal panel. Before making changes, inspect the available inputs and tell me what is missing. Work on one representative example first. Explain your proposed approach, preserve links to the source data, and send any write or external action to review before applying it.

Typical estimate for one setup or run. Data volume, complexity, and required human review can change the total time.

Where to find it

Agent conversation → Terminal panel

How to get started

  1. 1 Open the Terminal panel in a conversation to follow live commands, output, working directory, and exit status while the agent works.
    • Where: Open Agent conversation → Terminal panel.
    • Prepare: Make sure the required source rows, files, or URLs are available, then Open the Terminal panel in a conversation to follow live commands, output, working directory, and exit status while the agent works.
    • Check: Start with one representative example so you can adjust the setup safely.

    Select the step again to collapse these instructions.

  2. 2 Take control at any time and type your own commands into the same Terminal Session; both you and the agent keep the same shell state, current directory, and environment.
    • Where: Continue in Agent conversation → Terminal panel.
    • Configure: Take control at any time and type your own commands into the same Terminal Session; both you and the agent keep the same shell state, current directory, and environment.
    • Check: Review every field, permission, and input, then use a small test case before applying it to all your data.

    Select the step again to collapse these instructions.

  3. 3 Select the Command Entries you want the agent to know about, attach them to your next message, and continue the conversation with the exact command and output included as context.
    • Where: Complete this step in Agent conversation → Terminal panel.
    • Run: Select the Command Entries you want the agent to know about, attach them to your next message, and continue the conversation with the exact command and output included as context.
    • Verify: Inspect the result and its source data, then approve or save it only when it matches your expectations.

    Select the step again to collapse these instructions.

Natural next action