llama-mcp

Public source for low-friction install — not an open-source product. Most operations require a Llama Ventures team account (tokens are minted by the team admin at /settings/tokens). The one exception is the public pitch family.

Install

npm i -g @llamaventures/cli    # Node 18+; also puts `llama-mcp` on your PATH
llama --version
llama auth status              # round-trips against /api/me

Related MCP server: Small Business MCP Server

Authenticate

Credentials are tried in this order, on every call:

llama auth login          # browser sign-in; tokens auto-refresh, survive reboots
llama auth logout         # revokes server-side, clears local storage
llama token set llc_…     # PAT from /settings/tokens — validated before it lands on disk
llama auth status         # shows the resolved identity + active method

No account? Ask your Llama Ventures contact — any email address can be granted a token.

Integrate your AI system

This package is the supported integration surface for Llama Command. Wire in-house agents and LLM apps through here — not the raw HTTP API: the CLI/MCP layer owns the auth chain, the stable Error[…] contract, and forward-compatibility (SemVer); raw API routes carry no such promise.

1. Credentials — llama auth login, or a PAT via llama token set / $LLAMA_TOKEN for headless systems.

2. Install — npm i -g @llamaventures/cli.

3. Wire it in — MCP-native agents point at llama-mcp (per-client config); anything else shells out to llama ….

4. Onboard the agent — run llama agent-onboard (or the MCP agent_briefing prompt) at session start. It returns the server-owned Agent Runtime Contract, always in sync with the live server.

5. Verify — llama auth status, then llama deal search "<anything>".

CLI

The CLI is the canonical interface — it handles auth, error formatting, and schema forward-compatibility. Prefer it even from scripts.

llama deal search "acme ai"            # find deals (deal list takes the same filters)
llama deal show <dealId>
llama deal feed <dealId>               # every contribution, newest first
llama activity new-deals --since 24h   # recent deal creations
llama activity updated-deals --since 7d # meaningful updates grouped by deal
llama deal create "Acme AI" --source alex --deal-owner owner@llamaventures.vc --source-direction Outbound --status Interested
llama deal fact add <dealId> --category funding --claim "Raised a seed round" --source "deck p3" --source-url https://...
llama deal update <dealId> status Diligence
llama post <dealId> "note body"
llama brief add-text <dealId> --heading "..." --body "..."
llama wiki search "<query>"
llama wiki save <slug> --title "..." --content "..."
llama mentions
llama agent-onboard                    # server-owned agent workflow contract

Status vocabulary — Interested: tracked before any contact · Outreached: contacted, no response yet · Sourced: real relationship signal exists. sourceDirection is separate: Inbound came to the firm, Outbound we reached out first.

Facts use claim for the fact text. source is a readable provenance label and sourceUrl is the canonical evidence URL; both round-trip from the API. For deal owners, use an exact /api/field-options dealOwner value, a user email, or a numeric user id.

Run llama --help for the group index, llama help all for the full reference (100+ commands). Deletes are soft and audit-logged everywhere.

Error codes

The MCP server returns the same prefixes inside isError: true content. Authenticated calls send bounded, content-redacted telemetry to Command; llama eval good|bad --last turns real searches into eval feedback.

MCP server

llama-mcp is a stdio Model Context Protocol server exposing 59 typed tools that mirror the most-used CLI surface. Every tool is named and scoped — there is deliberately no generic API passthrough. Auth is identical to the CLI's chain. For the exact live list, pipe tools/list through it:

printf '%s\n' \
  '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"dev","version":"1"}}}' \
  '{"jsonrpc":"2.0","method":"notifications/initialized"}' \
  '{"jsonrpc":"2.0","id":2,"method":"tools/list"}' \
  | llama-mcp

~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{ "mcpServers": { "llama": { "command": "llama-mcp" } } }

claude mcp add llama -- llama-mcp

Point the client at the llama-mcp binary (which llama-mcp). Same JSON shape as above; no protocol extensions, no transport flags.

New agent? llama agent-onboard (CLI) or the agent_briefing prompt (MCP) pulls the server-owned workflow contract. The bundled AGENT_BRIEFING.md is a fallback copy only.

External pitch — no Llama account required

Founders, EAs, and prospective hires can pitch without a token: the pitch commands (and pitch_* MCP tools) talk to our public intake agent — the same structured 12-dimension intake as the web version, driven from your terminal or your own AI agent.

llama pitch start --name "Jane Doe" --email "jane@acme.ai"
llama pitch say "We're building an AI dev tool for X..."
llama pitch upload ./deck.pdf
llama pitch                     # interactive REPL

Server-enforced rate limits apply (per-IP, per-email, per-session).

Stability

• SemVer. Renaming/removing a command → major; new tool/command/flag → minor; fixes → patch.

• Public contract: the wire format (Bearer / X-Llama-Token) and the Error[…] prefixes don't change inside a major version.

• No raw-API passthrough, by design. If a wrapper you need hasn't landed, open an issue instead of calling the HTTP API directly.

Security

• Published via npm Trusted Publishers (OIDC) with --provenance — no npm token exists to leak.

• Zero runtime deps for the CLI; the MCP server depends only on @modelcontextprotocol/sdk, pinned exact.

• Branch protection, Dependabot, secret scanning, push protection enabled.

• Tokens: ~/.llama/token mode 0600 locally; sha256 hashes server-side.

Report vulnerabilities privately via GitHub security advisories — not public issues. See SECURITY.md.

Contributing

Internal tool maintained by Llama Ventures; team PRs welcome — see CONTRIBUTING.md. External: issues for doc gaps and broken flows are welcome; to get your company in front of us, use the pitch path.

License

MIT — © 2026 Llama Ventures, Inc.