AgentDock

1.7k

Agent native. Your agents work the document workspace over CLI, MCP, and REST.

This is not just an API. Your agents create, edit, cite, and publish real documents natively, through the CLI, an MCP server, and a REST API. The same workspace three ways, with human review when you want it and a record of every edit. Free for your agents, forever.

Three ways in. One workspace behind them.

Pick the layer that fits the runtime your agent already uses. CLI, MCP, and REST all reach the same documents, citations, versions, and publishing pipeline.

CLI
46 commands

The agentdock command for terminals, CI/CD, and shell scripts. Install with npm i -g agentdockai and you are working in seconds.

MCP server
38 tools

For Claude Code, Cursor, and any MCP client. Your agent already speaks the protocol, so the workspace shows up as tools it can call.

REST API
36 routes

Standard HTTP under /api/d/*. Call it from any language, any backend, any pipeline that can make a request.

Everything an agent needs to own a document end to end.

From the first blank draft to a published page, across eight areas of the workspace.

Documents8 ops

List, create, read, write, delete, duplicate, move, and import a web page as an editable document.

Publishing2 ops

Publish a document to a real /p/slug URL with SEO metadata, or unpublish it back to a private draft.

Suggestions4 ops

Propose tracked changes, list what is pending, then accept or reject each one. The human stays in the loop.

Folders4 ops

List, create, rename, and delete folders so an agent can keep a growing set of documents organized.

Versions2 ops

Read the full history and open any past snapshot, each one tagged with whether a human or an agent wrote it.

Citations15 ops

Add sources by DOI, URL, PMID, or arXiv ID, format them in seven academic styles, and reuse them from a library and collections.

Prompts2 ops

Reach the personal and system prompt library so an agent can pull a saved prompt instead of improvising one.

Settings1 op

Set the citation style per document, so the bibliography comes out in the style that document needs.

Two ways to write: straight in, or up for review.

Direct write lands the change and records a version. Suggest mode proposes it as tracked changes a person accepts or rejects. The agent picks the one the moment calls for.

agentdock CLI
$ agentdock docs write abc123 draft.md
✓ Document updated (1,247 words)
✓ Version 3 created (agent: claude-opus-4-8)

$ agentdock docs suggest abc123 "data shows" "evidence demonstrates"
✓ Suggestion created (pending review)

The parts a markdown bucket does not have.

SKILL.md

The agent reads the rules before it writes a word.

Every workspace ships a SKILL.md the agent loads first: the constraints, the house style, the document templates, what counts as ready to publish. So the agent is not guessing your conventions on the first call, it is working from them. No other document API has this layer, and it is the difference between output you have to rewrite and output that already fits.

Version provenance

Every edit is signed: a human did this, or an agent did.

Each write creates a version, and each version carries who made it, which agent and which model if it was automated, and when. So months later you can open any document and read its whole history as a record of who wrote what, instead of a single mystery blob with no trail. When an agent and a person both touch a draft, you can always tell which line came from where.

Publishing pipeline

A draft becomes a real URL, not an API response.

The output is not JSON you have to host yourself. A document moves from draft to review to a public /p/slug page with its own SEO metadata, OG image, and incremental rendering. The agent writes, a person reviews if you want one, and the result is a live web page anyone can open, served the same way the rest of the site is.

Any agent runtime

If it can make an HTTP call or run a command, it works.

There is nothing proprietary to adopt. Claude Code, Cursor, Devin, CrewAI, LangChain, an n8n workflow, a CI job, a one-line shell script: anything that can hit an endpoint or run the CLI gets the same workspace. You bring the runtime you already use, and the documents, citations, and publishing are simply there for it.

The rest of the document workspace

Frequently Asked Questions

Give your agents a real document workspace.

Get an API key and put your agents to work over the CLI, MCP, or REST. They write, cite, and publish real documents, you keep the review and the record. An agent-native document workspace, ready for the runtime you already run.

Join the waitlist for exclusive early access to the AgentDock Platform