Skip to main content
Glama
adrojis

tracecat-mcp-community

by adrojis
README.md
<p align="center">
  <img src="assets/banner.png" alt="Tracecat MCP Banner" width="100%">
</p>

# tracecat-mcp-community

**A full-stack Model Context Protocol (MCP) server for the [Tracecat](https://tracecat.com) SOAR platform — 94 tools across 16 domains.**

[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
[![npm version](https://img.shields.io/npm/v/tracecat-mcp-community.svg)](https://www.npmjs.com/package/tracecat-mcp-community)
[![Node.js](https://img.shields.io/badge/node-%3E%3D18-brightgreen.svg)](https://nodejs.org)
[![CI](https://github.com/adrojis/tracecat-mcp-community/actions/workflows/ci.yml/badge.svg)](https://github.com/adrojis/tracecat-mcp-community/actions/workflows/ci.yml)
[![Tracecat](https://img.shields.io/badge/Tracecat-v1.0.0--beta.42-purple.svg)](https://github.com/TracecatHQ/tracecat)
[![MCP](https://img.shields.io/badge/MCP-Server-green.svg)](https://modelcontextprotocol.io)
[![Docker](https://img.shields.io/badge/Docker-ready-blue.svg)](#docker)

---

## What is this?

An [MCP server](https://modelcontextprotocol.io) that gives AI assistants (Claude Code, Claude Desktop, etc.) full control over a [Tracecat](https://github.com/TracecatHQ/tracecat) instance through natural language. Manage workflows, actions, cases, secrets, tables, schedules, graphs, and more — without leaving your editor.

- **94 tools** covering the full Tracecat API surface, including write operations not exposed by the official MCP
- **Stdio transport** — no OIDC/SSO setup required, works against any self-hosted Tracecat
- **Lazy authentication** — MCP transport starts instantly, login happens on first tool call
- **Auto workspace detection** — no manual workspace ID needed
- **Session cookie auth** — handles Tracecat's cookie-based auth transparently

---

## Community MCP vs Official MCP

Tracecat ships an [official MCP server](https://docs.tracecat.com/automations/tracecat-mcp) (HTTP transport, OIDC auth, bundled with the platform). This community MCP is a **standalone alternative** you can use instead — pick the one that fits your setup.

| | Community MCP (this project) | [Official MCP](https://docs.tracecat.com/automations/tracecat-mcp) |
|---|---|---|
| **Transport** | stdio (local) | HTTP (remote) |
| **Auth** | Session cookie (username/password) | OIDC / SSO |
| **Setup** | `npx tracecat-mcp-community` + `.env` | Requires OIDC configured on the Tracecat instance |
| **Tool coverage** | 94 tools — full CRUD + graph ops + autofix + variables/integrations | ~90 tools — read + basic CRUD, plus agents/skills (EE-oriented) |
| **Exclusive capabilities** | Graph editing (`add_edges`, `move_nodes`, `autofix_workflow`), Schedules CRUD, Secrets write, Cases CRUD + tasks + fields, Actions CRUD, Variables CRUD, Integrations management, Templates, Webhook key rotation | Agent presets/skills/sessions (Enterprise Edition), official support, platform-integrated |
| **Best for** | Local dev, self-hosted without SSO, workflow authoring/editing at scale | Teams already running Tracecat Cloud or self-hosted with OIDC |

You don't need both. This community MCP is designed to cover the full surface on its own.

---

## Tools

| Domain | Tools | Description |
|---|---|---|
| **Workflows** | 9 | List, create, get, update, deploy, export, delete, validate, autofix |
| **Actions** | 5 | List, create, get, update, delete workflow actions |
| **Executions** | 6 | Run workflows, run drafts, list/get/cancel executions, compact view |
| **Cases** | 15 | List, create, get, update, delete cases; comments; tasks CRUD; custom fields CRUD |
| **Secrets** | 5 | Search, create, get, update, delete secrets |
| **Variables** | 6 | List, search, get, create, update, delete non-sensitive workspace variables |
| **Tables** | 5 | List, create, get, update, delete tables |
| **Columns** | 2 | Create, delete table columns |
| **Rows** | 6 | List, get, insert, update, delete, batch insert rows |
| **Schedules** | 5 | List, create, get, update, delete schedules |
| **Graph** | 5 | Get graph, add/delete edges, move nodes, update trigger position |
| **Folders** | 5 | List, create, update, delete folders; move workflows into folders |
| **Workspaces** | 5 | Get current, list, create, update, delete workspaces |
| **Integrations** | 7 | List/get/test/disconnect/delete integrations; list/get OAuth providers |
| **Webhooks** | 3 | Get/update webhook, rotate API keys |
| **Docs** | 2 | Search Tracecat docs, list available tool documentation |
| **Templates** | 2 | List and get community workflow templates |
| **System** | 1 | Health check |

> **Total: 94 tools** for complete Tracecat automation.

---

## Quick Start

### Option A: npx (fastest)

```bash
# Install globally
npm install -g tracecat-mcp-community
```

Create a `.env` file wherever you run from (or in the package directory):

```env
TRACECAT_API_URL=http://localhost/api
TRACECAT_USERNAME=your-email@example.com
TRACECAT_PASSWORD=your-password-here
TRACECAT_WORKSPACE_ID=              # Optional — auto-detected if omitted
```

Add to your `.mcp.json`:

```json
{
  "mcpServers": {
    "tracecat": {
      "command": "npx",
      "args": ["-y", "tracecat-mcp-community"]
    }
  }
}
```

### Option B: From source

```bash
git clone https://github.com/adrojis/tracecat-mcp-community.git
cd tracecat-mcp-community
npm install
cp .env.example .env    # Edit with your credentials
npm run build
```

Add to your `.mcp.json`:

```json
{
  "mcpServers": {
    "tracecat": {
      "command": "node",
      "args": ["/absolute/path/to/tracecat-mcp-community/dist/index.js"]
    }
  }
}
```

### Option C: Docker

```bash
git clone https://github.com/adrojis/tracecat-mcp-community.git
cd tracecat-mcp-community
docker build -t tracecat-mcp-community .
```

```json
{
  "mcpServers": {
    "tracecat": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "--env-file", "/path/to/.env", "tracecat-mcp-community"]
    }
  }
}
```

> **Security:** `.env` is gitignored and never committed. Never hardcode credentials in source files. See [SECURITY.md](SECURITY.md).

Then restart Claude Code and verify with `/mcp` — you should see the `tracecat` server with 94 tools.

---

## Configuration

| Variable | Required | Default | Description |
|---|---|---|---|
| `TRACECAT_API_URL` | No | `http://localhost/api` | Tracecat API base URL |
| `TRACECAT_USERNAME` | **Yes** | — | Login email |
| `TRACECAT_PASSWORD` | **Yes** | — | Login password |
| `TRACECAT_WORKSPACE_ID` | No | Auto-detected | Workspace ID (uses first workspace if omitted) |

Credentials are loaded from `.env` via [dotenv](https://github.com/motdotla/dotenv). The `.env` file must be in the project root (next to `package.json`).

---

## Architecture

```
src/
├── index.ts          # Entry point — StdioTransport + env loading
├── server.ts         # McpServer creation + tool registration
├── client.ts         # HTTP client with lazy auth + auto workspace injection
├── types.ts          # TypeScript interfaces
└── tools/
    ├── workflows.ts    # Workflow CRUD + deploy/export/validate/autofix
    ├── actions.ts      # Action CRUD with YAML inputs
    ├── cases.ts        # Case CRUD + comments + tasks + custom fields
    ├── executions.ts   # Run (live + draft), list, cancel, inspect executions
    ├── secrets.ts      # Secret management
    ├── variables.ts    # Non-sensitive workspace variables CRUD
    ├── tables.ts       # Tables, columns, and rows
    ├── graph.ts        # Graph operations (get graph, edges, node positions)
    ├── folders.ts      # Folder CRUD + move workflow into folder
    ├── workspaces.ts   # Workspace CRUD + current-workspace info
    ├── integrations.ts # OAuth integrations + providers
    ├── webhooks.ts     # Webhook get/update + key rotation
    ├── schedules.ts    # Cron/interval scheduling
    ├── docs.ts         # Documentation search
    ├── templates.ts    # Community workflow templates
    └── system.ts       # Health check
```

---

## Key Design Decisions

| Decision | Rationale |
|---|---|
| **Lazy initialization** | MCP transport starts immediately; login happens on first tool call. Avoids blocking Claude Code startup. |
| **Session cookies** | Tracecat currently uses `fastapiusersauth` cookies, not API keys. The client handles login and cookie extraction automatically. See note below on upcoming API token support. |
| **YAML string inputs** | Action `inputs` are sent as YAML strings per the Tracecat API contract, not JSON objects. |
| **POST for updates** | Actions, secrets, and schedules use `POST` for updates instead of the conventional `PATCH`. |
| **Auto workspace injection** | `workspace_id` is auto-detected and injected as a query parameter on every request. |
| **Optimistic locking** | Graph operations read `base_version` before patching to prevent concurrent edit conflicts. |

---

## Authentication Roadmap

This server currently authenticates via **username/password** (session cookies). The Tracecat team is actively working on **API token authentication**, which will provide a simpler and more secure connection method — no more password in `.env`.

We will add API token support as soon as it becomes available upstream. The username/password method will remain supported for backward compatibility.

---

## API Quirks

These behaviors differ from typical REST conventions and are handled transparently by the server:

| Quirk | Details |
|---|---|
| `workspace_id` as query param | Must be `?workspace_id=...`, not a header |
| POST for updates | `/actions/{id}`, `/secrets/{id}`, `/schedules/{id}` use POST |
| Actions list endpoint | `GET /actions?workflow_id=...` (not nested under `/workflows`) |
| Action inputs format | YAML string, not JSON object |
| Workflow list pagination | Returns `{ items: [...], next_cursor }`, not a plain array |

---

## Development

```bash
# Watch mode (auto-reload)
npm run dev

# Build TypeScript
npm run build

# Run directly
node dist/index.js
```

## Testing

```bash
npm run build
npm test
```

Tests use Node.js built-in test runner (no extra dependencies). See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.

## MCP Inspector

The [MCP Inspector](https://github.com/modelcontextprotocol/inspector) is a visual debugging tool that lets you browse and test all 94 tools interactively in your browser — useful for verifying your setup, exploring tool schemas, and testing API calls without Claude.

From the project root:

```bash
npx @modelcontextprotocol/inspector node dist/index.js
```

This starts a local web UI (default: `http://localhost:6274`). Click the **Tools** tab to see all available tools, inspect their input schemas, and execute them against your Tracecat instance.

---

## Roadmap

This project is under active development. Tracecat's API surface evolves fast, and we intend to keep up — expect new tools, refinements, and breaking-change adaptations as the platform matures.

Planned areas of improvement:

- **More tools** — covering new Tracecat API endpoints as they ship
- **Better error handling** — structured error responses with actionable hints
- **OAuth/OIDC support** — for Tracecat instances using SSO instead of basic auth
- **Test suite** — automated integration tests against a live Tracecat instance

Contributions, issues, and feature requests are welcome.

---

## Related Projects

- [Tracecat official MCP](https://docs.tracecat.com/automations/tracecat-mcp) — HTTP/OIDC MCP bundled with the platform (alternative to this one)
- [tracecat-skills](https://github.com/adrojis/tracecat-skills) — Claude Code skills for Tracecat workflow building
- [Tracecat](https://github.com/TracecatHQ/tracecat) — The open-source SOAR platform
- [MCP SDK](https://github.com/modelcontextprotocol/typescript-sdk) — Model Context Protocol TypeScript SDK

> **Note:** This project was previously named `tracecat-mcp`. It was renamed to `tracecat-mcp-community` in April 2026 to distinguish it from Tracecat's official MCP server (HTTP + OIDC), which shipped shortly after.

---

## License

[MIT](LICENSE)

TDQS

B3.2/5.0

Scored across 94 tools

Disambiguation4/5

Most tools have distinct purposes with clear descriptions, but some pairs like tracecat_get_execution / tracecat_get_execution_compact and tracecat_run_workflow / tracecat_run_workflow_draft could cause confusion. Overall, the delineation is good.

Naming Consistency5/5

All tools follow the consistent 'tracecat_verb_noun' pattern (e.g., tracecat_list_cases, tracecat_delete_workflow). Verbs and nouns are uniformly used, making the surface predictable.

Tool Count1/5

With 94 tools, the count is extremely high for any MCP server. While the coverage is broad, this number likely overwhelms agents and exceeds the typical well-scoped range of 3–15 tools.

Completeness5/5

The tool set covers CRUD operations for all major entities (workflows, cases, tables, secrets, integrations, etc.) and includes validation, debugging, and deployment. No obvious gaps in lifecycle coverage.

Maintenance

ActivityInactive
ResponsivenessNo issues