Quickstart

Prerequisites

• Rust toolchain — install via rustup if you don’t have it
• Git — with submodule support

1. Clone and Build

git clone --recurse-submodules https://github.com/jacobprall/moneypenny.git
cd moneypenny
cargo build --release

Put the binary on your PATH:

cp target/release/mp /usr/local/bin/mp

Or use the setup script, which also downloads the local embedding model:

./scripts/setup.sh

2. Initialize

mp init

This creates:

• moneypenny.toml — project configuration
• mp-data/ — data directory with one SQLite file per agent
• A default agent named main

3. Connect to Your Agent

Moneypenny exposes its full surface area as an MCP server. One command registers it with your agent runtime — no manual JSON editing required.

Claude Code

mp setup claude-code

Writes .mcp.json in your project root. For global registration across all projects:

mp setup claude-code --global

Restart Claude Code to pick up the new MCP server.

Cortex Code CLI

mp setup cortex

Runs cortex mcp add to register directly. Verify with:

cortex mcp list

OpenClaw

mp setup openclaw

Writes to ~/.clawdbot/clawdbot.json. Restart OpenClaw to pick it up.

Verify

Start a conversation and ask:

What Moneypenny tools do you have?

Your agent should list the available MCP tools. Once connected, every operation below can be performed by asking your agent in plain English. The following steps show both ways.

4. Add Some Knowledge

Ingest a document:

CLI

mp ingest path/to/document.md

Natural Language

Ask your MCP-connected agent:

Ingest the file at path/to/document.md

Or from a URL:

CLI

mp ingest --url "https://example.com/docs"

Natural Language

Ask your MCP-connected agent:

Ingest the docs page at https://example.com/docs

5. Add Facts

Store structured knowledge:

CLI

echo '{"op":"memory.fact.add","args":{"content":"Our API rate limit is 1000 req/min per tenant.","summary":"API rate limit: 1000 req/min/tenant","pointer":"API: rate-limit","confidence":0.95,"keywords":"api rate limit tenant"}}' \
  | mp sidecar

Natural Language

Ask your MCP-connected agent:

Remember that our API rate limit is 1000 requests per minute per tenant.

List what was stored:

CLI

mp facts list

Natural Language

Ask your MCP-connected agent:

List all the facts you know.

6. Search

CLI

mp facts search "rate limit"
mp knowledge search "deployment"

Natural Language

Ask your MCP-connected agent:

What do you know about rate limits?

Search the knowledge base for anything about deployment.

7. Chat (Requires LLM)

For interactive conversation with your agent:

CLI

mp chat

Natural Language

If you’re already in an MCP-connected client, you’re chatting! Just talk:

What do you know about our deployment pipeline?

Note

Chat and send commands require an LLM provider. The default config uses Anthropic as the generation provider. To use a local model or different API provider, edit moneypenny.toml and set provider = "local" or provider = "http" with your API base URL and key.

For a one-shot message:

CLI

mp send main "What do you know about our deployment pipeline?"

Natural Language

Ask your MCP-connected agent:

Ask the main agent what it knows about our deployment pipeline.

8. Start the Gateway

For multi-agent, multi-channel operation:

CLI

mp start

Natural Language

Ask your MCP-connected agent:

Start the Moneypenny gateway.

This spawns worker processes for each agent, starts the scheduler, and binds configured channel adapters (HTTP, Slack, Discord, Telegram).

What’s Next

Agents How agents work and how to manage them.

Memory & Facts The structured memory system.

Policy Engine Govern what your agents can do.

Ingestion Guide Import documents, URLs, and external event histories.