Skip to main content

A local-first AI agent that runs a business on NOSTR: checks its Lightning wallet, publishes notes and listings, answers encrypted DMs, and pays for paywalled MCP tools over NWC — all under sats budget caps it can't exceed. Drives five MCP servers (nwc-mcp, nostr-ops-mcp, marketplace-mcp, albyhub-admin-mcp, paywall-mcp). Ollama-first, API-pluggable. Built on pydantic-ai — no LangChain. MIT.

Project description

nostr-merchant

The AI that runs your NOSTR business. A local-first Python agent that holds the whole merchant toolkit — it checks its Lightning wallet, publishes notes and listings, runs a Shopstr storefront, answers encrypted DMs, and pays for paywalled MCP tools over NWC — all under sats budget caps it can't exceed. Built on five MCP servers (nwc-mcp, nostr-ops-mcp, marketplace-mcp, albyhub-admin-mcp, paywall-mcp) — 37 tools the agent picks from.

Ollama-first, API-pluggable. Built on pydantic-ai — no LangChain. MIT.

v0.2 — renamed from llmops-agent. All six CLI commands wired. The agent layer enforces a budget-and-audit safety pipeline on top of every MCP server's own safety stack. This release ships the full toolkit, the safety stack, and a working self-paying loop as proof the agent-pays-for-tools path holds end to end (given a working LLM backend — Ollama / Anthropic / OpenAI). Directing it at your own merchant tasks is what it's for today; unattended scheduled storefront-tending is the roadmap.


Install

pipx install nostr-merchant          # recommended: isolated CLI
# or:  uvx --from nostr-merchant nostr-merchant --help
# or:  pip install nostr-merchant

The five MCP servers it drives are launched on demand via npx -y — no separate install. You need Python 3.11+, an NWC-compatible wallet (e.g. Alby Hub), and an LLM backend (see Usage).


Status

Layer State
Design doc ✅ Done — ../03-python-reference-agent-design.md
Package scaffold (uv init, pyproject.toml, ruff/mypy/pytest config) ✅ Done
config.py — env-driven Pydantic Settings ✅ Done
budget.py — persistent rolling-window tracker ✅ Done
audit.py — NDJSON writer ✅ Done
mcp_servers.pyMCPServerStdio launch + doctor probe ✅ Done
agent.py — Pydantic AI agent + process_tool_call middleware ✅ Done
workflows/research.py — self-paying research workflow ✅ Done
cli.py — typer entry-point with ask / doctor / budget / audit / config-print / version ✅ Done
Live LLM smoke test (Anthropic Haiku) ✅ Done — end-to-end ask returns a real nwc_get_balance receipt with all 37 substrate tools loaded

Usage

# from this directory, with uv-installed venv
uv run nostr-merchant --help
uv run nostr-merchant version
uv run nostr-merchant budget                # snapshot, no LLM needed
uv run nostr-merchant config-print          # effective config, secrets masked
uv run nostr-merchant doctor                # ping each configured MCP server
uv run nostr-merchant ask "What's my Lightning wallet balance?"   # the main demo loop
uv run nostr-merchant audit --tail 20       # recent audit entries

Or run the whole sequence with the bundled script: ./smoke.sh (full run — the ask is read-only, so it can never move sats) or ./smoke.sh --no-ask (skip the LLM call entirely).

ask requires a working LLM backend, set via NOSTR_MERCHANT_MODEL:

  • anthropic:claude-haiku-4-5-20251001 (or any Anthropic/OpenAI model) — fast, reliable tool-calling. Set ANTHROPIC_API_KEY (or OPENAI_API_KEY) in ~/.nostr-merchant/.env. Recommended for most setups.
  • ollama:<model> — fully local, no KYC, no phone-home (the sovereignty default). Needs Ollama at localhost:11434. Reality check: a 37-tool agent loop wants real hardware — a small model on a CPU-only box is too slow to be practical. Run Ollama on a GPU, or use a hosted model and keep your wallet + keys local (the trust boundary that actually matters).

Provider creds in ~/.nostr-merchant/.env are loaded automatically — no need to export them.


Dev loop

uv sync                 # install deps + dev deps into .venv
uv run ruff check       # lint
uv run mypy             # type check
uv run pytest           # tests

Build wheel for distribution:

uv build

License

MIT — see LICENSE.

Contact

Built by LLMOps.Pro.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nostr_merchant-0.2.0.tar.gz (20.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

nostr_merchant-0.2.0-py3-none-any.whl (25.1 kB view details)

Uploaded Python 3

File details

Details for the file nostr_merchant-0.2.0.tar.gz.

File metadata

  • Download URL: nostr_merchant-0.2.0.tar.gz
  • Upload date:
  • Size: 20.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.12

File hashes

Hashes for nostr_merchant-0.2.0.tar.gz
Algorithm Hash digest
SHA256 ffa4ea69ae51ec664895b3455de652902441524946a498765080580e71736613
MD5 2c186930a1ea376745597d283e69eea7
BLAKE2b-256 448cc83132ae244f4603e2e5d281a84aaabc6d3a69f0871d27c9018298176c96

See more details on using hashes here.

File details

Details for the file nostr_merchant-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: nostr_merchant-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 25.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.12

File hashes

Hashes for nostr_merchant-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 e5ca6d1ce3117f2fd900aa511019f8760ef299086fe53f6c4bbde3a3b4b534e3
MD5 4a4e1f4e88b622d6996a9b985705ed9f
BLAKE2b-256 3ff8597236c8b51edae22bcc6262e0b5be99c680bde08747c573fd644fadbbf5

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page