evc-mesh-mcp

# EVC Mesh MCP Server

<!-- mcp-name: io.github.entire-vc/evc-mesh-mcp -->

[![Install via Spark](https://spark.entire.vc/badges/evc-mesh-mcp/install.svg)](https://spark.entire.vc/assets/evc-mesh-mcp?utm_source=github&utm_medium=readme)

[Model Context Protocol](https://modelcontextprotocol.io/) (MCP) server for [EVC Mesh](https://github.com/entire-vc/evc-mesh) — a task management platform for coordinating humans and AI agents.

Connects AI agents (Claude Code, Cursor, Cline, OpenClaw, etc.) to EVC Mesh via MCP tools for task management, persistent memory, event publishing, and multi-agent coordination.

**This is the actively developed copy.** [`evc-mesh`](https://github.com/entire-vc/evc-mesh) also ships an MCP server (`./cmd/mcp`, same `internal/mcp` tool set) that it builds and deploys itself — the two exist because Go's `internal/` visibility rules mean one repo can't import the other's package, not because they're meant to diverge. New tools and fixes land here first.

## Prerequisites

- Go 1.22+
- Running EVC Mesh instance
- Agent registered in Mesh with an API key (`agk_...`)

## Installation

```bash
go install github.com/entire-vc/evc-mesh-mcp@latest
```

Or build from source:

```bash
git clone https://github.com/entire-vc/evc-mesh-mcp.git
cd evc-mesh-mcp
go build -o evc-mesh-mcp .
```

### Docker

```bash
docker run -i --rm \
  -e MESH_API_URL \
  -e MESH_AGENT_KEY \
  ghcr.io/entire-vc/evc-mesh-mcp
```

`-i` is required — the server speaks MCP over stdio, and Docker only wires up
stdin when the container runs interactively. Add `-e MESH_MCP_PROFILE=core`
to switch profiles (see [Tool Profiles](#tool-profiles) below). The image is
published for `linux/amd64` and `linux/arm64` from `Dockerfile` in this repo
on every tagged release (`docs/RELEASING.md`).

## Tool Profiles

The MCP server supports two profiles to optimize context window usage:

| Profile | Tools | Context overhead | Best for |
|---------|-------|-----------------|----------|
| **core** | 25 | ~8K tokens (4% of 200K) | Claude Code, Cursor, small-context models |
| **full** | 63 | ~18K tokens (9% of 200K) | Power users, automation agents, admin ops |

Set via `MESH_MCP_PROFILE` environment variable. Default: `full`.

## Configuration

| Variable | Required | Default | Description |
|----------|----------|---------|-------------|
| `MESH_API_URL` | Yes | `http://localhost:8005` | Base URL of the Mesh API |
| `MESH_AGENT_KEY` | Yes (stdio) | — | Agent API key (`agk_...`) |
| `MESH_MCP_PROFILE` | No | `full` | Tool profile for stdio: `core` or `full` (SSE serves both) |
| `MESH_MCP_TRANSPORT` | No | `stdio` | Transport mode: `stdio` or `sse` |
| `MESH_MCP_HOST` | No | `0.0.0.0` | SSE server bind host |
| `MESH_MCP_PORT` | No | `8081` | SSE server bind port |
| `MESH_MCP_AUTH_FAIL_RPM` | No | `20` | SSE mode: per-IP budget for authentication attempts against a not-yet-cached agent key on `/sse`, `/core/sse`, `/mcp`, `/mcp/core`. Over budget → `429` without calling Mesh API. `0` disables. |
| `MESH_MCP_SESSION_CACHE_TTL_MIN` | No | `15` | SSE mode: how long a successful authentication is trusted before the key is re-checked — bounds how long a revoked key keeps working without a restart. |
| `MESH_MCP_AUTH_FAIL_CACHE_SEC` | No | `30` | SSE mode: how long a failed authentication (bad/unknown key) is remembered, so repeating the same bad key doesn't call Mesh API every request. |
| `MESH_MCP_DECIDER_USERNAME` | No | — | Username recorded as `decided_by` when `record_owner_decision` answers a gated task. Unset: the workspace owner. |
| `MESH_MCP_LEGACY_TOOL_ALIASES` | No | off | `1` also registers the tools' earlier names, for deployments whose callers still use them. Leave off for new installs. |

### Running without credentials

In stdio mode the server also starts when `MESH_AGENT_KEY` is not set. It then
answers `initialize` and `tools/list` as usual, and every tool call returns
instructions for setting `MESH_API_URL` and `MESH_AGENT_KEY`. This lets MCP
clients and catalogs inspect the tool list before you have a key. If a key is
set but authentication fails at startup (API unreachable, key rejected), the
server keeps running: tools are listed, and each call retries authentication
and returns the reason until it succeeds.

### Tool annotations

Every tool declares the MCP hints `readOnlyHint`, `destructiveHint`,
`idempotentHint` and `openWorldHint`, so clients can tell read-only tools
(`get_*`, `list_*`, `recall`, `search_docs`, …) from ones that change or
remove data (`update_*`, `move_task`, `forget`, …).

### Client metrics

Each `initialize` is logged with the client's `clientInfo.name` and version,
and counted in the Prometheus metric `mesh_mcp_initialize_total{client,profile}`
(exposed on `/metrics` in SSE mode; client names are normalised and capped).

### Claude Code (stdio mode)

Add to your project's `.mcp.json`:

```json
{
  "mcpServers": {
    "evc-mesh": {
      "command": "evc-mesh-mcp",
      "env": {
        "MESH_API_URL": "https://your-mesh-instance.example.com",
        "MESH_AGENT_KEY": "agk_your-workspace_your-key",
        "MESH_MCP_PROFILE": "core"
      }
    }
  }
}
```

### Cursor

Add to Cursor MCP settings (Settings → MCP Servers):

```json
{
  "evc-mesh": {
    "command": "evc-mesh-mcp",
    "env": {
      "MESH_API_URL": "https://your-mesh-instance.example.com",
      "MESH_AGENT_KEY": "agk_your-workspace_your-key",
      "MESH_MCP_PROFILE": "core"
    }
  }
}
```

### SSE Mode (multi-agent, shared server)

For connecting multiple agents through a shared MCP endpoint:

```bash
MESH_API_URL=https://your-mesh-instance.example.com \
MESH_MCP_PORT=8081 \
evc-mesh-mcp --transport sse
```

SSE mode serves **two profiles simultaneously** on different paths:

| Path | Profile | Description |
|------|---------|-------------|
| `/sse` + `/message` | full | All 63 tools (backward compatible) |
| `/core/sse` + `/core/message` | core | 25 essential tools |

The same process also serves the **Streamable HTTP** transport (stateless, one
agent key per request, sent in the `Authorization: Bearer` or `X-Agent-Key`
header — the query parameter is refused there):

| Path | Profile |
|------|---------|
| `/mcp` | full |
| `/core` | core |

Authentication per connection via:
- `Authorization: Bearer agk_...` header
- `X-Agent-Key: agk_...` header
- `?agent_key=agk_...` query parameter

## Agent Context Protocol (ACP)

At session start, follow these 5 steps in order:

```
1. heartbeat(status="online")              → register as alive
2. get_project_knowledge(project_id)       → load accumulated decisions & conventions
3. get_my_rules(project_id)                → understand constraints
4. get_context(project_id)                 → see recent activity + project knowledge
5. get_my_tasks()                          → check assigned work
```

At session end:
```
publish_event(type="summary", memory={persist: true})  → broadcast + persist
session_report(model, tokens_in, tokens_out)           → report metrics
```

## MCP Tools — Core Profile (25)

### ACP & Identity

| Tool | Description |
|------|-------------|
| `heartbeat` | Send heartbeat. Call at session start with status=online. Response includes `mesh_version` (the running binary's build git-SHA, or `"dev"` for an unpinned local build) — cheap way to check whether a fix has actually reached the installed binary without shelling out to the host. |
| `get_project_knowledge` | Get ALL permanent knowledge (decisions, conventions). ACP Step 2 |
| `get_my_rules` | Get ALL governance rules (workflow + assignment). ACP Step 3 |
| `get_context` | Get recent activity + project knowledge. ACP Step 4 |
| `get_my_tasks` | Get assigned tasks. ACP Step 5 |

### Task Management

| Tool | Description |
|------|-------------|
| `list_projects` | List workspace projects |
| `list_tasks` | List tasks with filters (status, priority, assignee, search) |
| `get_task` | Get task details with optional comments/artifacts/deps |
| `create_task` | Create a new task |
| `update_task` | Update task fields |
| `move_task` | Change task status using slugs |
| `assign_task` | Assign/unassign a task |
| `get_task_context` | Get everything about a task in one call |
| `add_vcs_link` | Link a task to a pull request, commit or branch |

### Communication

| Tool | Description |
|------|-------------|
| `add_comment` | Add comment to a task (markdown). Response includes a `delivery` array per `@`-mention reporting whether it actually reached the recipient (task queue/notification) or was skipped/failed and why |
| `publish_event` | Publish event + optional memory hint for persistence |

### Memory

| Tool | Description |
|------|-------------|
| `recall` | Search memory by keywords |
| `remember` | Save knowledge (UPSERT by key) |
| `forget` | Delete a memory entry |
| `recall_with_graph` | Search memory, expanding results through the knowledge graph |
| `set_project_knowledge` | Write a structured project fact (upsert by key) |
| `get_canonical_updates` | Fetch canonical decisions recorded since a given time |
| `record_owner_decision` | Record a decision by the workspace owner as canonical project knowledge |

#### What `recall` guarantees about its result

**`limit` is a hard bound.** The response never contains more than `limit` items,
and `total` always equals the number of items actually returned. Nothing is added
to the page after it has been sized — not pinned rows, not graph-expanded
neighbours.

**Rows that fail `scope`/`tags`/`tags_any` are dropped, never returned unmarked.**
This holds regardless of how a row reached the result: ordinary retrieval, pinning,
or graph expansion. A pinned row is exempt from *ranking*, not from *eligibility* —
"pinned" means "do not let ranking bury this", not "show this to a caller who asked
for a different scope".

**Graph neighbours are marked and bounded.** With `RECALL_GRAPH_ENABLED=true`,
`recall` also runs a knowledge-graph expansion and folds in `hop > 0` neighbours,
ea