# Co-Create MCP server

> Developer reference. Endpoint: https://co-create.build/api/mcp. Updated 2026-11-11. Web version: https://co-create.build/developers/mcp

<!-- Generated from src/lib/mcp/mcpDocs.ts. Edit that file, then run: npx tsx scripts/write-mcp-docs.ts -->

Connect your Co-Create account to Claude Code, Claude Desktop, Cursor, VS Code or any other MCP client, so an AI assistant can read your projects, add to your plans and check on your builds.

The Co-Create MCP server speaks the [Model Context Protocol](https://modelcontextprotocol.io) over Streamable HTTP at `https://co-create.build/api/mcp`. You sign in with a personal access token you make in Co-Create. Everything the assistant does is done as you, with exactly the access you have in the app.

## Quick start with Claude Code

1. In Co-Create, open Settings, AI & building, Connect to Claude Code (MCP). Name the token (for example "Claude Code on my laptop"), choose Read and write, and tap Make token. Copy it: it is shown once.
2. In a terminal, run the command below with your token in place of the example.
3. Start Claude Code and run `/mcp`. `co-create` is listed as connected with its tools. Ask something like "List my Co-Create projects".

```bash
claude mcp add --transport http co-create https://co-create.build/api/mcp --header "Authorization: Bearer cc_pat_YOUR_TOKEN"
```

Add `--scope user` to use it in every folder. Do not combine `--scope project` with a token in the command: that writes the header, token included, into the folder's shared `.mcp.json`. To share the server with a team, use the environment variable form under Setting up each client.

## Endpoint

|  | Value |
| --- | --- |
| URL | `https://co-create.build/api/mcp` |
| Transport | Streamable HTTP. POST one JSON-RPC 2.0 message, or a batch of up to 20, per request. Every answer is a single `application/json` body. |
| Protocol versions | `2025-06-18`, `2025-03-26`, `2024-11-05`. `initialize` answers with the version you asked for when it is one of these, otherwise the newest. |
| Sessions | None. The server is stateless and sends no `Mcp-Session-Id`. GET (an event stream) and DELETE answer 405. |
| Authentication | `Authorization: Bearer cc_pat_...` on every request. |
| Capabilities | `tools` only (no resources, prompts or sampling). |
| Server | `co-create` version 1.0.0 |

## Authentication

A token looks like `cc_pat_` followed by 43 letters, digits, `-` and `_`. Send it as a bearer token on every request. There is no OAuth flow: a client that offers to "authenticate" this server in a browser does not need to; give it the header instead.

- **Read and write** tokens can use every tool. **Read only** tokens can use the read tools, and `tools/list` shows them only those.
- A token expires after the time you chose when you made it (30 days, 90 days, 365 days, never). An expired or revoked token answers 401.
- You can have 10 active tokens. Make one per device or app, so you can revoke one without the others.
- Making a token needs you to be fully signed in to Co-Create (including two-step sign-in, if you use it). Using one needs only the token, so keep it like a password.

## Making and revoking a token

1. Open Co-Create and go to Settings, AI & building, Connect to Claude Code (MCP).
2. Type a name, choose Read and write or Read only, and choose when it expires.
3. Tap Make token and copy it. Co-Create keeps only a fingerprint of it, so it cannot be shown again; if you lose it, make another.
4. To stop a token working, tap Revoke beside it. It stops at the next request.

The list shows each token's first characters, when it was made, when it was last used and when it expires.

## Setting up each client

### Claude Code

```bash
claude mcp add --transport http co-create https://co-create.build/api/mcp --header "Authorization: Bearer cc_pat_YOUR_TOKEN"
```

Or add it to a folder's `.mcp.json` and keep the token in an environment variable (Claude Code expands `${VAR}`):

```json
{
  "mcpServers": {
    "co-create": {
      "type": "http",
      "url": "https://co-create.build/api/mcp",
      "headers": {
        "Authorization": "Bearer ${CO_CREATE_TOKEN}"
      }
    }
  }
}
```

Check it with `claude mcp list` in a terminal, or `/mcp` inside Claude Code. Remove it with `claude mcp remove co-create`.

### Claude Desktop

Claude Desktop's own connector screen expects OAuth, so connect through the `mcp-remote` bridge, which adds the header for you. Edit `claude_desktop_config.json` (Settings, Developer, Edit Config), then restart Claude Desktop. Needs Node.js 18 or newer.

```json
{
  "mcpServers": {
    "co-create": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://co-create.build/api/mcp",
        "--header",
        "Authorization:${CO_CREATE_AUTH}"
      ],
      "env": {
        "CO_CREATE_AUTH": "Bearer cc_pat_YOUR_TOKEN"
      }
    }
  }
}
```

The token rides in `env` because some systems split arguments that contain spaces.

### Cursor

Add this to `~/.cursor/mcp.json` (every project) or `.cursor/mcp.json` (one project):

```json
{
  "mcpServers": {
    "co-create": {
      "url": "https://co-create.build/api/mcp",
      "headers": {
        "Authorization": "Bearer cc_pat_YOUR_TOKEN"
      }
    }
  }
}
```

### VS Code (GitHub Copilot)

Add this to `.vscode/mcp.json`. VS Code asks for the token once and stores it securely:

```json
{
  "inputs": [
    {
      "type": "promptString",
      "id": "co-create-token",
      "description": "Co-Create MCP token",
      "password": true
    }
  ],
  "servers": {
    "co-create": {
      "type": "http",
      "url": "https://co-create.build/api/mcp",
      "headers": {
        "Authorization": "Bearer ${input:co-create-token}"
      }
    }
  }
}
```

### Any other MCP client

Use the Streamable HTTP (sometimes called "HTTP" or "remote") transport, the URL `https://co-create.build/api/mcp`, and the header `Authorization: Bearer cc_pat_YOUR_TOKEN`. A client that only speaks stdio can use `npx -y mcp-remote` as in the Claude Desktop example.

### Raw HTTP

Each call is one POST. With your token in `CO_CREATE_TOKEN`:

```bash
curl -s https://co-create.build/api/mcp \
  -H "Authorization: Bearer $CO_CREATE_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"curl","version":"1"}}}'
```

```bash
curl -s https://co-create.build/api/mcp \
  -H "Authorization: Bearer $CO_CREATE_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'
```

```bash
curl -s https://co-create.build/api/mcp \
  -H "Authorization: Bearer $CO_CREATE_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"list_projects","arguments":{"limit":5}}}'
```

A tool result has the JSON in `content[0].text` and the same object in `structuredContent`:

```json
{
  "jsonrpc": "2.0",
  "id": 3,
  "result": {
    "content": [
      {
        "type": "text",
        "text": "{\n  \"projects\": [ ... ]\n}"
      }
    ],
    "structuredContent": {
      "projects": [
        "..."
      ]
    },
    "isError": false
  }
}
```

## Tools

Every tool acts as the token's account and sees exactly what that account sees in Co-Create: projects it owns or collaborates on, chats it is a member of, apps it owns. No tool takes a user or account argument. Writes only ever add to a project's Slop, Stacks or Output, or reword one Stacks bullet, exactly as if you had typed it; collaborators with the project open see the change arrive.

| Tool | What it does | Token |
| --- | --- | --- |
| `whoami` | Who am I | Read only or read and write |
| `list_projects` | List projects | Read only or read and write |
| `read_project` | Read a project | Read only or read and write |
| `add_to_slop` | Add to Slop | Read and write |
| `add_stacks_bullets` | Add bullets to Stacks | Read and write |
| `edit_stacks_bullet` | Edit a Stacks bullet | Read and write |
| `add_to_output` | Add to Output | Read and write |
| `list_coding_runs` | List coding runs | Read only or read and write |
| `get_coding_run` | Get a coding run | Read only or read and write |
| `list_apps` | List apps | Read only or read and write |
| `list_chats` | List chats | Read only or read and write |
| `read_chat` | Read a chat | Read only or read and write |

### whoami

**Who am I.** The Co-Create account this token acts as, and what the token may do (read, or read and write).

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "whoami",
  "arguments": {}
}
```

```json
{
  "userId": "3f6c...",
  "name": "Dana Lee",
  "email": "dana@example.com",
  "tokenName": "Claude Code on my laptop",
  "scope": "read_write"
}
```

### list_projects

**List projects.** The projects this account owns or is a collaborator on, most recently updated first. Archived projects are left out. Each project has an id the other tools take, a link that opens it in Co-Create, and the app it builds into, if any.

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "maxLength": 200,
      "description": "Only projects whose name contains this text (case-insensitive)."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 100,
      "default": 50,
      "description": "At most this many projects."
    }
  },
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "list_projects",
  "arguments": {
    "query": "recipe"
  }
}
```

```json
{
  "projects": [
    {
      "projectId": "8b1d...",
      "name": "Recipe Box",
      "isOwner": true,
      "updatedAt": "2026-11-11T18:02:11.000Z",
      "app": {
        "appId": "c41a...",
        "name": "Recipe Box"
      },
      "url": "https://co-create.build/projects/8b1d..."
    }
  ]
}
```

### read_project

**Read a project.** One tab of a project's document. `slop` is the freeform notes, `stacks` is the plan as headers and bullets (each bullet shown as `- [bullet id] text`, the id edit_stacks_bullet takes), `output` is the written draft. Long tabs come in pages: pass `nextOffset` back as `offset` to read on. `revision` is the document's revision at the time of the read.

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The project's id, from list_projects."
    },
    "tab": {
      "type": "string",
      "enum": [
        "slop",
        "stacks",
        "output"
      ],
      "default": "stacks",
      "description": "Which tab to read."
    },
    "section": {
      "type": "string",
      "maxLength": 400,
      "description": "Stacks only: just the header with this exact title (case-insensitive)."
    },
    "offset": {
      "type": "integer",
      "minimum": 0,
      "default": 0,
      "description": "Character offset to start from."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 50000,
      "default": 20000,
      "description": "At most this many characters."
    }
  },
  "required": [
    "projectId"
  ],
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "read_project",
  "arguments": {
    "projectId": "8b1d...",
    "tab": "stacks"
  }
}
```

```json
{
  "projectId": "8b1d...",
  "name": "Recipe Box",
  "tab": "stacks",
  "revision": 42,
  "text": "## Sign in\n- [b-0k2j1a-1x9c0d] Sign in with a magic link\n\n## Recipes\n- [b-3jd0q1-9a8s7d] Add a recipe from a photo",
  "offset": 0,
  "nextOffset": null,
  "charsTotal": 112,
  "truncated": false,
  "url": "https://co-create.build/projects/8b1d..."
}
```

### add_to_slop

**Add to Slop.** Appends text to the end of a project's Slop (its freeform notes), the same as typing it there. Collaborators with the project open see it arrive.

Needs a read and write token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The project's id, from list_projects."
    },
    "text": {
      "type": "string",
      "minLength": 1,
      "maxLength": 20000,
      "description": "What to add. Markdown-style headers and `- ` bullets are kept as written."
    }
  },
  "required": [
    "projectId",
    "text"
  ],
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "add_to_slop",
  "arguments": {
    "projectId": "8b1d...",
    "text": "- Let people share a recipe by link"
  }
}
```

```json
{
  "ok": true,
  "summary": "Added a line to Slop.",
  "revision": 43,
  "url": "https://co-create.build/projects/8b1d...?focus=..."
}
```

### add_stacks_bullets

**Add bullets to Stacks.** Adds bullets under a Stacks header, creating the header at the end when no header has that title. Indent a bullet by two spaces per level to nest it. Nothing is approved or built: bullets are added exactly as if typed.

Needs a read and write token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The project's id, from list_projects."
    },
    "header": {
      "type": "string",
      "minLength": 1,
      "maxLength": 400,
      "description": "The header's title. An existing header is matched case-insensitively."
    },
    "bullets": {
      "type": "array",
      "minItems": 1,
      "maxItems": 20,
      "items": {
        "type": "string",
        "minLength": 1,
        "maxLength": 2000
      },
      "description": "The bullets, without a leading dash."
    }
  },
  "required": [
    "projectId",
    "header",
    "bullets"
  ],
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "add_stacks_bullets",
  "arguments": {
    "projectId": "8b1d...",
    "header": "Recipes",
    "bullets": [
      "Scale a recipe up or down",
      "Print a shopping list"
    ]
  }
}
```

```json
{
  "ok": true,
  "summary": "Added 2 bullets under \"Recipes\".",
  "revision": 44,
  "url": "https://co-create.build/projects/8b1d...?focus=..."
}
```

### edit_stacks_bullet

**Edit a Stacks bullet.** Rewrites one Stacks bullet, named by the id read_project shows. A bullet that has been approved, built, prepped, answered or has screenshots is refused (worked_bullets_need_confirm): change it in Co-Create, which asks first.

Needs a read and write token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The project's id, from list_projects."
    },
    "bulletId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 40,
      "description": "The `b-...` id from read_project."
    },
    "text": {
      "type": "string",
      "minLength": 1,
      "maxLength": 2000,
      "description": "The bullet's new text."
    },
    "expectedRevision": {
      "type": "integer",
      "minimum": 0,
      "description": "Optional: refuse (revision_conflict) unless the document is still at this revision."
    }
  },
  "required": [
    "projectId",
    "bulletId",
    "text"
  ],
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "edit_stacks_bullet",
  "arguments": {
    "projectId": "8b1d...",
    "bulletId": "b-3jd0q1-9a8s7d",
    "text": "Add a recipe from a photo or a link"
  }
}
```

```json
{
  "ok": true,
  "summary": "Updated a bullet in \"Recipes\".",
  "revision": 45,
  "url": "https://co-create.build/projects/8b1d...?focus=..."
}
```

### add_to_output

**Add to Output.** Appends a paragraph or more to the end of a project's Output (its written draft).

Needs a read and write token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The project's id, from list_projects."
    },
    "text": {
      "type": "string",
      "minLength": 1,
      "maxLength": 20000,
      "description": "What to add."
    }
  },
  "required": [
    "projectId",
    "text"
  ],
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "add_to_output",
  "arguments": {
    "projectId": "8b1d...",
    "text": "Launch checklist: invite five friends, then post in the group."
  }
}
```

```json
{
  "ok": true,
  "summary": "Added to the Output draft.",
  "revision": 46,
  "url": "https://co-create.build/projects/8b1d...?focus=..."
}
```

### list_coding_runs

**List coding runs.** Recent coding runs (Remy building an approved bullet), newest first: across every project this account is on, or one project's. Status is one of queued, kicked_off, coding, done, failed.

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "Only this project's runs."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 100,
      "default": 20,
      "description": "At most this many runs."
    }
  },
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "list_coding_runs",
  "arguments": {
    "limit": 1
  }
}
```

```json
{
  "runs": [
    {
      "runId": "5e0f...",
      "projectId": "8b1d...",
      "header": "Recipes",
      "bullet": "Add a recipe from a photo",
      "status": "done",
      "currentStep": "Merged",
      "summary": "Added photo import with on-device text recognition.",
      "startedAt": "2026-11-11T17:40:00.000Z",
      "finishedAt": "2026-11-11T18:01:30.000Z",
      "startedBy": "Dana Lee"
    }
  ]
}
```

### get_coding_run

**Get a coding run.** One coding run in full: its status, current step, summary, and any questions it is waiting on you to answer (answer them in Co-Create).

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "runId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The run's id, from list_coding_runs."
    }
  },
  "required": [
    "runId"
  ],
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "get_coding_run",
  "arguments": {
    "runId": "5e0f..."
  }
}
```

```json
{
  "runId": "5e0f...",
  "status": "coding",
  "currentStep": "Writing the photo import screen",
  "pendingQuestions": [],
  "url": "https://co-create.build/projects/8b1d..."
}
```

### list_apps

**List apps.** The apps this account owns, with their status, preview link, and Live link when the app is published.

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "list_apps",
  "arguments": {}
}
```

```json
{
  "apps": [
    {
      "appId": "c41a...",
      "name": "Recipe Box",
      "status": "ready",
      "projectId": "8b1d...",
      "previewUrl": "https://recipe-box-preview.example.com",
      "liveUrl": "https://recipebox.example.com",
      "liveStatus": "live"
    }
  ]
}
```

### list_chats

**List chats.** The conversations this account is in: project chats, group chats, direct messages and Remy threads. Each has a chatId read_chat takes.

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "list_chats",
  "arguments": {}
}
```

```json
{
  "chats": [
    {
      "chatId": "9c2e...",
      "name": "Recipe Box",
      "kind": "project",
      "projectId": "8b1d..."
    }
  ]
}
```

### read_chat

**Read a chat.** Messages in one conversation, newest first. Name the chat by `chatId`, or by `projectId` for that project's chat. Pass `nextCursor` back as `before` for older messages. Only messages this account can see in Co-Create are returned.

Works with a read-only token.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "chatId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "The chat's id, from list_chats."
    },
    "projectId": {
      "type": "string",
      "minLength": 1,
      "maxLength": 100,
      "description": "Read this project's chat instead of naming a chatId."
    },
    "before": {
      "type": "string",
      "maxLength": 200,
      "description": "A nextCursor from an earlier call."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 100,
      "default": 30,
      "description": "At most this many messages."
    }
  },
  "additionalProperties": false
}
```

Example arguments and result:

```json
{
  "name": "read_chat",
  "arguments": {
    "projectId": "8b1d...",
    "limit": 2
  }
}
```

```json
{
  "chatId": "9c2e...",
  "messages": [
    {
      "messageId": "m2...",
      "from": "Remy",
      "fromKind": "assistant",
      "text": "Photo import is live in your preview.",
      "at": "2026-11-11T18:02:00.000Z"
    },
    {
      "messageId": "m1...",
      "from": "Dana Lee",
      "fromKind": "human",
      "text": "Can recipes come from a photo?",
      "at": "2026-11-11T17:39:12.000Z"
    }
  ],
  "nextCursor": "2026-11-11T17:39:12.000Z|m1..."
}
```

## Errors

Problems with the request itself are HTTP statuses with a JSON-RPC error body:

| HTTP | JSON-RPC code | Meaning |
| --- | --- | --- |
| 400 | -32700 | The body is not valid JSON. |
| 400 | -32600 | Not a valid JSON-RPC request, an empty or oversized batch, or an unsupported `MCP-Protocol-Version` header. |
| 401 | -32001 | No token, or a token that is unknown, revoked or expired. `error.data.reason` says which; the `WWW-Authenticate` header carries it too. |
| 405 | -32600 | GET or DELETE: this server has no event stream and no sessions. |
| 413 | -32600 | The body is larger than 1 MB. |
| 429 | -32002 or -32001 | Too many requests; wait for `Retry-After` seconds. |
| 503 | -32603 | Co-Create could not check the token just now; retry shortly. |
| 200 | -32601 | Unknown method. |
| 200 | -32602 | Unknown tool name. |

A tool that runs but cannot do what was asked answers HTTP 200 with `isError: true`, and `structuredContent` is `{ "error": "<code>", "message": "<what to do>" }`, so the assistant can read it and correct itself. The codes:

| error | Meaning |
| --- | --- |
| `invalid_arguments` | An argument is missing, of the wrong type, out of range, or not an argument of this tool. |
| `read_only_token` | A write tool was called with a read-only token. |
| `project_not_found` | No such project, or this account is not on it (the two are not told apart). |
| `section_not_found`, `section_ambiguous` | read_project's `section` matched no header, or more than one. |
| `target_not_found`, `target_ambiguous` | edit_stacks_bullet's id matched no bullet (it may have been edited), or more than one. |
| `worked_bullets_need_confirm` | The edit would change a bullet with work on it; make it in Co-Create, which asks first. |
| `revision_conflict` | `expectedRevision` no longer matches; read again and retry. |
| `conflict` | Someone else saved at the same moment several times running; retry. |
| `run_not_found`, `chat_not_found` | No such run or chat for this account. |
| `internal_error` | Something failed on Co-Create's side; retry, and contact us if it keeps happening. |

## Limits

- At most 20 messages and 1 MB per request.
- Per token: 120 requests a minute that call tools, 30 of them writes. Over that is HTTP 429 with `Retry-After: 60`.
- 30 requests with a bad token per network per 15 minutes, then HTTP 429.
- Every tool answers in well under the platform's 26 second limit: none of them waits on an AI model or a build.
- Reads are paged: read_project returns up to 50,000 characters at a time (`nextOffset` continues), read_chat up to 100 messages (`nextCursor` continues), the list tools up to 100 items.
- Text written by add_to_slop or add_to_output is at most 20,000 characters per call; add_stacks_bullets takes up to 20 bullets of 2,000 characters each.

## Security

- A token is a password for your account's projects. Never paste it into a chat, a public repository or a shared `.mcp.json`; use an environment variable or your client's secret storage.
- Co-Create stores only a SHA-256 fingerprint of each token, never the token.
- Revoke a token in Settings the moment you stop using a device or suspect it leaked. Changing your password does not revoke tokens; revoke them too.
- Tokens stop working at once if your account is scheduled for deletion, and are deleted with it.
- Choose Read only for anything that only needs to look.
- Every tool call is recorded in Co-Create's logs with the tool, the outcome and the token (never the arguments).
- What an assistant reads through this server is your data; an assistant can be steered by text it reads (prompt injection), so review writes it proposes to make.

## What v1 leaves out

These need a person in Co-Create, either because they spend credits or because they have checks that happen on screen:

- Sending chat messages. A message can start a Remy reply, which uses your AI allowance. read_chat reads them; reply in Co-Create.
- Approving bullets and starting, stopping or answering coding runs (these spend credits and go through the approve screen).
- Creating, renaming, archiving, deleting, duplicating or sharing projects, and inviting people.
- Removing anything from a document, or changing a bullet that has work on it.
- Publishing to Live, domains, app settings and billing.
- Private Remy replies (cloud memory). In read_chat they appear as a placeholder; open the chat in Co-Create to read them.
- MD files (projects made from a markdown file) are not listed.

## For AI assistants setting this up

If you are an AI agent configuring this server for a person, do this:

1. Ask the person to make a token in Co-Create (Settings, AI & building, Connect to Claude Code (MCP)) and give it to you. Never guess, reuse or log one.
2. Claude Code: run `claude mcp add --transport http co-create https://co-create.build/api/mcp --header "Authorization: Bearer <token>"`. Other clients: transport Streamable HTTP, URL `https://co-create.build/api/mcp`, header `Authorization: Bearer <token>`.
3. Verify by calling `whoami`; it returns the account's name and the token's scope.
4. Call `list_projects` before any project tool; ids come from there. Call `read_project` with `tab: "stacks"` before `edit_stacks_bullet`; bullet ids come from there.
5. Treat `isError: true` results as answers: read `message` and follow it rather than retrying unchanged.

This reference is also served as plain markdown at https://co-create.build/developers/mcp.md, and listed in https://co-create.build/llms.txt. The live tool list and schemas are what `tools/list` returns.

## Changelog

- **1.0.0** (2026-11-11): first release. Twelve tools: whoami, list_projects, read_project, add_to_slop, add_stacks_bullets, edit_stacks_bullet, add_to_output, list_coding_runs, get_coding_run, list_apps, list_chats, read_chat.
