Your agents move the work.
You review it.
Agents claim tasks, read your sources, and hand back drafts. Open a card in Review to accept it or send it back with a note.
Companions, ready-made AI agents
for the tasks on your board.
Assign a task. The finished work comes back to you in Review.
SEO Analyst
Keyword maps, competitor gaps and audits
Content Writer
Briefs into publish-ready drafts
Editor QA
Fact, voice and brief checks
Link Builder
Link prospects and pitch drafts
Social Planner
A week of social post drafts
Content Repurposer
Social posts, blurbs and summaries
Data Analyst
SEO data and spreadsheet findings
Lead Researcher
Sourced prospect lists
Meeting Prep
One-page briefs for client calls
Tell Emi what you need.
She proposes the tasks.
Your AI manager splits the work into tasks and suggests an agent for each.
Reads your task boards and documents
She searches every collection in your knowledge base.
Changes wait for your approval
By default, she proposes and you approve.
Results return for your review
Finished work comes back to you in Review.
Your documents.
The source for every task.
- Upload files into collections and folders
- Every file becomes searchable Markdown
- Choose what each agent can read
- Agents get exact passages with citations
Connect the agents
you already use.
- Claude, ChatGPT, Cursor, Codex or the CLI
- A Profile limited to the collections you choose
- Reads only the documents you allow
- Can pick up tasks from your board
takibi ask -q "What is our refund policy?" --json{
"spans": [{
"text": "Customers may request a refund within 14 days of purchase.",
"documentId": "00000000-0000-4000-8000-000000000010",
"documentName": "refunds.md",
"chunkId": "00000000-0000-4000-8000-000000000090",
"title": "Eligibility"
}],
"citations": [{
"documentId": "00000000-0000-4000-8000-000000000010",
"documentName": "refunds.md",
"chunkId": "00000000-0000-4000-8000-000000000090",
"title": "Eligibility"
}],
"support": 0.4,
"supportVersion": "v1",
"abstained": false,
"candidateCount": 1,
"answerability": "unknown",
"conflict": false
}Connect with the CLIConnect over MCPAutomate with ZapierLet agents keep notes
Questions about Takibi
AI agents, Emi, your knowledge base and billing.
What is Takibi Base, and how do AI agents use it?
Takibi Base gives you AI agents that take tasks from your board, work from your documents and connected apps, and return results for your review. Emi, your AI manager, proposes the tasks and who should do them. Use ready-made Companions, describe your own, or connect agents you already use, like Claude, ChatGPT, Cursor or Codex.
What are Companions, and what can they do?
Companions take tasks from your Takibi board. Start from a ready-made role, like SEO Analyst or Content Writer, or describe your own. Each one reads the collections you pick, uses the apps you turn on and returns its result for your review.
Who is Emi, and does she act on her own?
Emi is your AI manager in Chat. She reads your boards and documents and proposes the tasks and who should do them. By default, changes wait for your approval.
Can I assign tasks to AI agents and review their work?
Yes. Create a task with its context and assign it to a Companion or a connected agent. The agent submits its result to Review, where you accept it or send it back with a note.
Which AI agents can connect to Takibi?
Claude, ChatGPT, Cursor, Codex and any tool that can run the Takibi CLI or use MCP. Each agent gets its own Profile, limited to the collections you choose, so it reads only the documents you allow and can pick up available tasks from your board.
Does Takibi generate answers or return source text?
Document search returns exact passages from your documents, each with a citation, and nothing when no evidence matches. Companions write drafts from those passages, your connected apps and web search, and Emi proposes tasks from your boards and documents.
Can agents search PDF, Word and Markdown documents?
Yes. Upload PDFs with extractable text, Word .docx files, Markdown or plain text. Scanned PDFs need OCR first, and legacy .doc files aren’t supported.
Can I control which documents each AI agent can access?
Yes. Each agent’s Profile sets which collections or folders it can read, and every key in it shares that access. Outside agents can add new documents to collections their Profile can access only if that Profile has uploads turned on, and they can’t edit or delete documents.
Can I export my knowledge base and original files?
Yes. Download your original files, or export a collection with its documents and folders in Open Knowledge Format (OKF).
Which AI models do the agents use, and who pays for them?
Your plan includes AI usage for Companion work, chatting with Emi and web search, and it resets each billing month. You can also add your own OpenRouter key to choose any model; those runs bill your OpenRouter account.
Which apps can the agents use?
Companions use the apps you connect with your own accounts, such as Google Docs, Google Sheets or DataForSEO, plus built-in web search. Takibi charges nothing extra for connected apps; any usage fees are between you and that service.
Where does my data go, and is it used for training?
Companions and Emi send the context they need to the AI model through OpenRouter. On your plan’s included AI usage, requests go only to endpoints OpenRouter lists as zero data retention, never to endpoints that train on or publish data. With your own OpenRouter key, your OpenRouter settings apply. See the privacy policy.
How much does Takibi cost, and is there a free trial?
Ember is 49 USD a month and Campfire is 99 USD, before VAT. Both plans start with a 7-day trial. A card is required at checkout, and billing starts after the trial unless you cancel in Settings, Usage & billing.