Machinepulse

Summarize the A2A agent registry — snapshot total and verified counts plus new and left agents over the last 7 days. Read-only aggregation of daily-collected observatory data; no side effects. Use when the user asks how the AI agent registry is growing.

Report collector health — coverage, per-source freshness and missed snapshot runs with reasons, separating recent misses from the whole recorded history; internal-only sources are excluded. Read-only aggregation of daily-collected observatory data; no side effects. Use when the user asks whether the observatory data is fresh or complete.

List tracked ecosystem events — without filters, counts per event type over the last 7 days plus the latest event of each type and a plain-word legend; set event_type or canonical_key for the full feed (limit/offset page through it), set since=YYYY-MM-DD to widen the window. Read-only aggregation of daily-collected observatory data; no side effects. For counts rather than the feed use the `get_*_summary` tools. Use when the user asks what changed recently in the ecosystem.

Fetch the entity card for any tracked type (MCP server, agent, HF model, package) by canonical_key or natural name — all linked facts, categories, HF model license tiers, how to connect, recent events; field captions in fields_note. Name lookup matches exact aliases and canonical keys, not free-form text. Read-only aggregation of daily-collected observatory data; no side effects. Package download days may be missing (npm/PyPI reporting delay); missing days are null with notes, never fake zeros. Use when the user asks about a specific server, agent, model or package.

Summarize x402 agent-payment volume (DefiLlama) — latest stored day and the 7-day window split by chains with non-zero volume, with the data lag stated. Read-only aggregation of daily-collected observatory data; no side effects. Use when the user asks about agent payments or x402 volume.

List repositories seen on GitHub Trending — last N days (1-100), newest first, with rank and stars gained today; limit/offset page through the list. Read-only aggregation of daily-collected observatory data; no side effects. For a project ranking by measured change use `get_leaderboard` instead. Use when the user asks what is trending on GitHub.

Rank tracked entities (servers, agents, models) by a project-level metric — project_momentum (score: middle of star/download/HF-derivative growth ranks) or trend_days (days on GitHub Trending) over the 7d window; limit/offset page through the full ranked list. Numbers and catalog facts only, no quality scores. Read-only aggregation of daily-collected observatory data; no side effects. Package download days may be missing (npm/PyPI reporting delay); missing days are null with notes, never fake zeros. Use when the user asks what is rising; for the daily GitHub Trending list use `get_github_trending`.

Summarize the Hugging Face model watchlist — model count, new arrivals, derivative net flow and the license tier mix. Read-only aggregation of daily-collected observatory data; no side effects. Use when the user asks about open LLMs and their licenses.

Fetch the observatory front page in one call — MCP servers, A2A agents, HF models, x402 money, top-5 growth with one catalog line each, new arrivals per category and data coverage, a few headline numbers each. Read-only aggregation of daily-collected observatory data; no side effects; money figures carry their stated data lag. Use when the user wants the state of the whole market in a single request; for per-metric values with formulas use `get_pulse`.

Summarize the MCP server ecosystem — unique servers, 7-day arrivals vs the previous 7 days, per-registry counts and new arrivals per category. Read-only aggregation of daily-collected observatory data; no side effects. Use when the user asks how big or how fast-growing the MCP ecosystem is.

Fetch daily download counts for an npm package — up to 100 days, with total, latest day and average over days that have data; days without collection are null with a note. Read-only aggregation of daily-collected observatory data; no side effects. Days inside the npm reporting delay are partial (shown, excluded from latest and average), never fake zeros. Use when the user asks how an npm package is growing; for side-by-side comparison use `get_package_comparison`.

Compare npm/PyPI download activity for up to 20 packages — last 30 days: total, latest day and average (a longer list is cut and says so). Read-only aggregation of daily-collected observatory data; no side effects. Package download days may be missing (npm/PyPI reporting delay); missing days are null with notes, never fake zeros. Use when the user wants to compare package popularity side by side; for a single package's history use `get_npm_trends` or `get_pypi_trends`.

List the protocol chronicle (MCP, A2A, AP2, x402) — releases and milestones, newest first, optionally filtered by protocol; limit/offset page through the entries. Read-only aggregation of daily-collected observatory data; no side effects. Use when the user asks when a protocol shipped something.

Snapshot every observatory metric for one day — window values with formula, formula_plain, basis, and per-metric history depth (`metric_history_from`); insufficient history is a fact, never a fake zero. Read-only; daily collection, package download windows may lag. Use when the user asks for the current state of the AI agent ecosystem; for a narrative overview use `get_market_overview`, for one domain use the `get_*_summary` tools.

Fetch daily download counts for a PyPI package — up to 100 days, with total, latest day and average over days that have data; days without collection are null with a note. Read-only aggregation of daily-collected observatory data; no side effects. Days inside the PyPI reporting delay are partial (shown, excluded from latest and average), never fake zeros. Use when the user asks how a PyPI package is growing; for side-by-side comparison use `get_package_comparison`.

Search MCP servers by canonical_key, name or description — substring match, not fuzzy search, ordered by stars when available; optional filters archived, has_repo, category, min_stars, pushed_before run in one request; limit/offset page through the matches. Read-only aggregation of daily-collected observatory data; no side effects. Name/keyword lookup over MCP servers only; for the card of a known entity use `get_entity_profile`. Use when the user gives a name, a task word (PDF, payment) or a filter and wants keys and facts.