# RCLL MCP server

> Provides self-hosted shared memory for AI agent fleets using hierarchical rooms, halls, and layers for scoped semantic retrieval.

- Repository: https://github.com/Holetron-lab/fleet-memory
- Page: https://hostedmcp.org/servers/holetron-lab/fleet-memory
- Transport: stdio
- Stars: 0
- License: MIT
- Latest version: 0.1.0

## Install

### Claude Code

```sh
claude mcp add fleet-memory -- npx -y fleet-memory-mcp
```

### Cursor (`~/.cursor/mcp.json`)

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

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "fleet-memory": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "fleet-memory-mcp"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"fleet-memory","type":"stdio","command":"npx","args":["-y","fleet-memory-mcp"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

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

### Codex

```sh
codex mcp add fleet-memory -- npx -y fleet-memory-mcp
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.fleet-memory]
command = "npx"
args = ["-y", "fleet-memory-mcp"]
```

### Gemini CLI

```sh
gemini mcp add -s user fleet-memory npx -- -y fleet-memory-mcp
```

### Gemini settings.json (`~/.gemini/settings.json`)

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

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `FLEET_BANK` | no | no | Default memory bank ID for retain/recall. The pre-rebrand variable MEMPALACE_BANK is still read as a fallback. Migrating from hindsight-mempalace-mcp? Set this to mempalace-main to keep reading your existing bank; see mcp-server/README.md. |
| `FLEET_URL` | no | no | Base URL of the fleet-memory backend (self-hosted). Defaults to http://127.0.0.1:5100. The pre-rebrand name HINDSIGHT_URL is still read as a fallback. |
| `HINDSIGHT_API_EMBEDDINGS_LOCAL_MODEL` | no | no |  |

## Tools (5)

- `memory_retain`: Save a fact, observation, or document to long-term memory with automatic room/hall classification. Rooms: auth, pipeline, schema, infrastructure, ui, api, deployment, monitoring, agent, general. Halls: fact, event, decision, preference, discovery, procedure, warning. Layers: L0=Identity (always loaded), L1=Critical, L2=Session (default), L3=Deep.
- `memory_recall`: Search long-term memory for relevant facts. Uses semantic search with optional room/hall scoping for significantly improved retrieval accuracy. Supports layer cascade (L0 results always prioritized).
- `memory_reflect`: Deep reasoning over memory — synthesizes facts, finds patterns, answers complex questions with citations. Use for analysis: "What patterns emerge from recent events?" or "Summarize everything about X."
- `memory_compress`: Create compressed memory summaries (closets) from stored facts. Groups memories by room+hall and creates AI-generated summaries with source pointers. Use when a topic has accumulated many facts.
- `memory_bridge`: Create a cross-bank memory bridge (tunnel) between two related memories in different banks. Relations: same_concept, depends_on, contradicts, extends. Use when concepts in separate banks are related.

## FAQ

### Does the memory retrieval process require an LLM?

No, the read path never invokes a language model, meaning recall costs only CPU and zero model tokens.

### Can I run RCLL without an LLM key?

Yes, by setting LLM_PROVIDER=none, though this limits the system to a hybrid vector-and-lexical chunk store without fact extraction or entity resolution.

### Is there a pre-built Docker image or npm package available?

There is no packaged release or published container image yet; you must build it from the source tree.
