A direct editing interface for people
Express ideas with headings, lists, tables and diagrams. Organize documents into knowledge bases and folders. Review agent edits in the browser and keep writing.
Documents as shared working state
Human–agent collaboration starts with one shared document.
Doco is document infrastructure for humans and AI agents. People use a browser-based rich-text editor; agents use MCP, CLI, Skills or REST APIs. Both read, edit and review the same documents, carrying knowledge beyond a single conversation.
A shared document layer
One document layer. Many agents. Ongoing collaboration.
A document can be working state that a team and its agents maintain together. Doco provides documents, knowledge bases, stable block addresses and real-time synchronization. You choose the agent and decide what it may read or change.
Explicit read and write interfaces for agents
Search and read through MCP tools, the terminal CLI or REST APIs. Target a paragraph by its stable block ID. The document interface is independent of the model you choose.
The same document state for both
Agent writes enter the same Yjs collaborative document used by the browser. Connected clients receive updates; version history supports review and rollback without uploading exported copies.
The collaboration contract
Address. Check. Write. Review.
An agent needs to identify its target, detect concurrent edits and verify the result of a write.
- 01
Search and read
Find documents inside an authorized knowledge base. Read the relevant content and its version. Keep document and block references so people can verify the evidence.
- 02
Address individual blocks
A stable block ID identifies the target paragraph through moves and reordering. Keep the scope of a change as small as one block.
- 03
Write with a version check
Send the version you read with If-Match. A version conflict returns 409: re-read, reconcile the intended change and retry instead of silently overwriting a concurrent edit.
- 04
Review and continue together
Inspect changes in the editor, create version snapshots or roll back when needed. Version checks protect against concurrent overwrites; people still need to review the content.
Four entry points
From interactive agents to automated workflows
MCP server
Give MCP-compatible agents Doco tools to read and maintain documents. Clients such as Claude Code and Cursor share the same document contract.
Choose your MCP clientCLI
Log in, search and edit from terminals or scripts. Add document maintenance to your existing development and automation workflows.
Explore the CLI and channelsAgent Skill
Give your agent Doco conventions and workflow guidance for reading, checking and updating documents in the right order.
See Skill installationREST API
Build a custom integration using an OpenAPI 3.1 interface. Control access with Bearer tokens and scopes; use block-level writes and batch transactions.
Read the REST API referenceCollaboration workflows
Keep documents inside the workflow
Long-term agent memory
Keep useful context, decisions and experience in readable documents. People can correct and organize them; future agents can read and cite them.
Explore inspectable agent memoryTeam wikis and project documents
People provide judgment and review. Agents help maintain explanations, summarize progress and update runbooks. The team reads and edits one shared knowledge base.
Explore a human–agent team wikiPersonal knowledge bases
Keep notes, sources and drafts in editable documents. Authorize an agent to organize them when useful, and retain a migration path through Markdown export.
Explore agent-assisted knowledge workChoices and boundaries
A general document layer with open workflows
Doco provides editing, storage, collaboration and read/write interfaces. Your chosen client or system provides agent reasoning, models and orchestration. Search and knowledge features follow the current API; they are not a RAG service for arbitrary external sources.
- Version checks detect concurrent edits; they cannot prove agent-generated content is correct. Grant write permissions when the workflow needs them.
- Local browser state supports offline editing. Agent API access and synchronization across devices require a network and server connection.
- The editor is open source; see the repository for self-hosting scope and SaaS differences. Markdown and knowledge-base ZIP exports provide a migration path.
Questions and answers
Understand the collaboration model
01How is Doco different from an AI document in a chat?
A chat can generate a draft; Doco provides a shared document that can be maintained over time. People edit directly, external agents read and write through authorized interfaces, and updates enter the same online document.
02Do I have to use a particular model or agent?
No. MCP-compatible clients can connect to the Doco MCP server. Custom programs can use the CLI or REST API. The integration guide lists clients with configuration examples.
03Can Doco be the document layer for my own product?
You can connect document and knowledge-base reads and writes to your product or workflow through the open API. Check authentication, scopes, rate limits and deployment requirements first. The API reference defines the supported interfaces.
04How can I move my data to another tool?
Export documents as Markdown, or package a knowledge base or folder as a ZIP. ZIP exports include supported attachments and rewrite image links to relative paths. Other tools may render custom rich-text nodes differently.