MCP server · APIs.io
APIs.io API catalog MCP server
One-click install for Cursor, VS Code, Claude, and 20+ other MCP clients, powered by API Commons MCP Install — visit install.apicommons.org for more information.
Endpoint
Hosted endpoint · transport http
Connect
This is a remote MCP server — point an MCP client at the endpoint URL. Clients with native remote support (Claude, Cursor, VS Code, …):
# mcp client config (e.g. .mcp.json / mcp settings)
{
"mcpServers": {
"apis-io": {
"url": "https://apis.io/mcp"
}
}
}
For stdio-only clients (older Claude Desktop, etc.), bridge with mcp-remote:
# Add to claude_desktop_config.json
{
"mcpServers": {
"apis-io": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://apis.io/mcp"]
}
}
}
If the server requires authentication, add the provider's token/header (e.g. an X-API-Key) per its docs. A quick reachability check:
curl -i -X POST https://apis.io/mcp \
-H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
Tools
apis_io_search— Federated overview of the catalog for a query — top matching APIs, providers and tags in one call.get_playground_apis— Curated APIs that are SAFE TO EXPERIMENT WITH — no key (or an instant, free, self-serve key), no cost, and no real-world side effects. Free, and deliberately so: this is where you start if you are newfind_apis—get_api—find_providers—get_provider—find_tags—get_tag—find_industries—get_industry—find_regions—get_region—find_countries— Browse the country dimension — the finest of the four place taxonomies, and the one a buyer usually names ("who is in Germany"). Peer to find_industries / find_regions / find_areas. Search by text; soget_country— One country: metadata, counts, and a top sample of member providers (with the total). Use find_providers?country=slug for the full ranked list, or view=full here.find_areas—get_area—find_extensions— Browse every OpenAPI `x-` specification extension harvested across the catalog, with the provenance the OpenAPI Initiative registry does not carry: who publishes it, in how many documents, where in thget_extension— One OpenAPI specification extension: its publishers, occurrence and document counts, where it appears, and the value shapes observed. CHECK is_api_evangelist BEFORE treating a key as provider-originatfind_ratings—get_provider_rating—find_cohorts— START HERE for market-level questions. Every scored population of providers in the catalog — tags, industries, regions, areas and VC portfolios — as one browsable index. Filter by kind, minimum populaget_cohort— One cohort: its identity, its population, and the full member roster with each provider's Kin Score and agent readiness. Understanding plan. The analysis parts (stats/rankings/scores/capabilities) neefind_tag_groups— START HERE for capability-map questions. The 1,500+ tag groups the catalog computes — sets of tags that demonstrably travel together across providers (Cards, Logistics, Wealth Management, Observabilitget_tag_group— One tag group: its member tags and the full provider roster resolved from the catalog, each with Kin Score and agent readiness. Filter the roster by industry, region, country or area. Free; the DISTRItag_group_tags— The member tags of one tag group, each joined to its catalog record — provider count, api count, tag quality band and composite. Use it to see what a group is actually made of before trusting the groucohort_stats— UNDERSTANDING — the DISTRIBUTION for a whole market: Kin Score mean/median/min/max/stdev, agent-readiness spread, band split, per-facet averages, and artifact adoption rates. This is the market-statscohort_rankings— UNDERSTANDING — the leaderboard for a market, on TWO axes: rank by Kin Score composite AND rank by agent readiness. The two orders differ, and the gap between a provider's two ranks is usually the stocohort_scores— UNDERSTANDING — facet-level scores for every member of a cohort, each with the cohort average and the delta against it. A 60 in governance means nothing until you know the market sits at 45; this is tfind_operations— UNDERSTANDING — CATALOG-WIDE operation search: which providers expose a path matching your terms. "Who has a POST /refunds" had no answer before this — tags are what a company says about itself, a patdeprecated_operations— UNDERSTANDING — every provider publishing at least one operation marked deprecated in its own OpenAPI, ranked by how many. The lifecycle view of the catalog: 8,297 deprecated operations across 681 procohort_history— UNDERSTANDING — the SET-LEVEL time series: how a whole market's Kin Score and agent readiness moved, build by build. There was a per-provider series and a whole-dataset licence and nothing between thecohort_failures— UNDERSTANDING — what a whole market gets WRONG: the agent-readiness checks its members most commonly fail, ranked by what fixing each would move the cohort mean. The inverse of gap_analysis, which saycohort_capabilities— UNDERSTANDING — what every member of a cohort actually publishes: artifact counts by type plus the agent-readiness dimensions each one satisfies. The capability matrix behind a market report. Large cocompare_cohorts— UNDERSTANDING — two markets side by side on one rubric: "is US payments further along than UK banking?" Returns both distributions plus the deltas on score, agent readiness, every facet and every adopget_provider_operations— Every operation a provider exposes, across all of its OpenAPIs — method, path, operationId, summary, which API it belongs to, and whether it is deprecated. The shortcut for "what can I actually call hget_provider_tools— Every MCP tool a provider ships, with the operation each one wraps and its auth. Reports provenance per tool: `first-party` means the provider publishes that server, `derived` means API Evangelist genget_provider_schema— Every JSON Schema a provider publishes — the data shapes their API works in, in one list, without walking their artifacts. Free.get_provider_facets— UNDERSTANDING — one provider's score broken into its facets (discoverability, contract quality, governance, operational transparency, developer ergonomics, commercial clarity) plus any regulatory regiget_agent_readiness— EXPLORE — the agent-readiness dimensions and standalone score for ONE provider (spec presence, MCP server, auth clarity, idempotency, error semantics, rate-limit signal, well-known catalog, consent idfind_agent_readiness— UNDERSTANDING — the agent-readiness leaderboard across the whole catalog. Rank providers by agent readiness, and filter to those that DO satisfy (has) or do NOT satisfy (missing) specific dimensions —agent_readiness_dimensions— UNDERSTANDING — diffusion: for each agent-readiness dimension, how many scored providers satisfy it, out of how many, as a share. The denominator every "the agent web is/is not here yet" claim needs afind_capabilities— START HERE for "what can a business do with these APIs". The business-capability model — 333 top-level capabilities across 28 industries — with how many catalog providers reach each one. Defaults to cget_capability— One business capability: its definition, its sub-capabilities, the external authorities it cites, and WHICH catalog providers reach it with their ratings. Accepts a slug or a BC id (e.g. "payments-carget_capability_edges— UNDERSTANDING — the evidence behind a capability. Every edge that lands on it: the provider, the tag, the exact OpenAPI file, a calibrated confidence, and the QUOTED FRAGMENT OF THE PROVIDER'S OWN CONget_provider_business_capabilities— UNDERSTANDING — what ONE vendor's published APIs let a business DO, rolled up by capability, with the tag-level evidence under each. Answers "what is this vendor actually for" from its own contracts rget_provider_capabilities— What one provider publishes, counted by artifact type, with the agent dimensions it satisfies. Free — it is a fact about their own public surface. NOTE: this is artifact counts, NOT business capabilitget_provider_evidence— How a provider's score was established, part by part: first-party (they published it), verified (we fetched and confirmed it), or derived (we inferred it). Free — the basis for a claim should never siget_rating_rubric—find_artifacts—compare_providers—gap_analysis—whats_changed—recommend_stack—get_provider_apis— Every API one provider publishes, as a page you can filter — the provider's own catalog rather than the whole network's. Prefer this over find_apis with a `providers` filter when you already know theget_provider_artifacts—get_api_artifacts—get_openapi—get_provider_onboarding—find_similar_providers—find_similar_apis—get_industry_leaders—get_region_leaders—get_country_leaders— UNDERSTANDING — Top-rated providers in one country. Membership is derived and partial, so read the ranking as "the best of what we have placed there", not a national league table.get_area_leaders—get_rating_history—find_rating_movers—story_leads—what_can_i_fix— INFLUENCE — the ranked, costed, gated punch list for one provider you own. Every agent-readiness check with points still on the table (exact — these are real per-check results, each with what satisfiereadiness_gates— INFLUENCE — what blocks the next agent-readiness band, GATES BEFORE POINTS. Some bands are unreachable on score alone: agent-native requires idempotency AND a stable error envelope regardless of how msimulate_fixes— INFLUENCE — project a set of fixes onto a provider's agent-readiness score and band before doing the work. Re-applies the band gate to the PROJECTED state, so it will tell you when buying points stillclaim_listing— INFLUENCE — ask to claim a listing as yours. Returns 202 with a token and the hosts it will be checked against: publish it as a DNS TXT record, or as a file at /.well-known/apis-io-challenge, and leavgenerate_artifact— INFLUENCE — REQUEST an artifact you are being marked down for not having: apis-json, agent-card, mcp (one tool per operation you already publish), arazzo, rules (a Spectral ruleset scoped to your ownsubmit_artifact— INFLUENCE — tell us about an artifact you publish, rather than waiting for us to find it. Returns 202; a person fetches it, checks it and re-runs the pipeline against it. Track it with check_status.correct_facts— INFLUENCE — propose corrections to what we say about you: name, description, url, industries, tags. Returns 202 and is applied by a person — an API that let an owner rewrite their own description unredispute_finding— INFLUENCE — "you say we lack X, here it is." Send the claim in words and a URL that evidences it. Returns 202; a person fetches your evidence, checks it against what we hold, and emails you either wayset_visibility— INFLUENCE — ask to be shown less, or not at all. `restricted` keeps a name, a description and a link to your own site, unrated and out of every ranked view; `delisted` removes you entirely. Returns 20watch_listing— INFLUENCE — be told when a listing you care about moves: score changes, band changes, or a check closing. Registration only — the scoring pipeline sends the mail, this records what you want to hear ablist_watches— INFLUENCE — the listings you are watching, with the events registered for each and when you started. The read half of watch_listing: call it before registering another, so you neither duplicate a watcunwatch_listing— INFLUENCE — stop watching a listing. Takes effect immediately and removes the registration outright; there is no paused state. Removing a watch that was never registered is not an error.request_check— INFLUENCE — ask APIs.io to (re)check something. Use it when a listing is wrong or stale, when a score looks incorrect, or to get a company profiled that is not in the catalog yet. A check can target acheck_status— INFLUENCE — the status of a check request by id: queued, in_review, running, done, rejected or needs_info, with any notes shared by the reviewer. The id identifies the request, but the Influence planreport_gap— FREE, no account — tell the catalog what you looked for and could not find. The consumer door (roadmap#166): a gap report is how the catalog learns what to add next, and a person reads the queue. Onemy_checks— INFLUENCE — the check requests you filed, newest first. Requires a key or login — an anonymous caller has no identity to scope a list to, and should read a single request with check_status instead.export_dataset—industry_gap_analysis—export_stack—insights_overview— UNDERSTANDING — the Insights root: how many companies are profiled, over how many quarters, and the top investment dimensions across the whole set. Read this FIRST — it tells you the size and vintagefind_company_insights—insights_dimensions—insights_adoption—insights_industries— UNDERSTANDING — the demand-side industry rollup: the sectors the profiled companies fall into. Each row carries TWO counts and they answer different questions: `company_count` is the research roster,get_company_insight—company_gaps—match_providers—find_vcs—get_vc—vc_portfolio—find_investors—resolve—enrich_provider—save_search—my_workspace— UNDERSTANDING — the workspace root: which plan you are on, how many saved searches and lists you hold, and where each lives. Answers "what do I have here and what can I afford" in one call, so an agenlist_saved_searches—run_saved_search—saved_search_net_new—delete_saved_search—create_list—list_lists—get_list—add_to_list—delete_list—find_openapis—find_asyncapis—find_channels—find_graphql—find_arazzo—find_mcp—find_skills—find_rules—find_scopes—find_security—find_plans—find_rate_limits—find_finops—find_collections—find_json_ld—find_json_schemas—find_json_structures—find_examples—find_postman—find_apis_json—submit_feedback— Report something wrong with a result: wrong data, a filter that did not filter, a ranking that put off-topic results first, a number you could not reproduce, or an answer that was useless for your tasreport_correction— Report that the catalog has a provider WRONG — a wrong website, a wrong logo, APIs that are not theirs, a count that does not match what they ship, a score computed from any of those. Free, always: co
About MCP
The Model Context Protocol (MCP) is an open protocol Anthropic introduced for connecting LLM-based agents to external tools and data sources. Providers publish MCP servers that expose their API surface as structured, discoverable tools — an MCP-compatible client (Claude Desktop, Cursor, Cline, Continue, etc.) can connect to the server and call its tools without any per-provider integration code.
Browse every MCP server on the APIs.io network or compare with the broader Agent Skill surfaces of the same providers.
Work with this as data
Every MCP server here is available over the APIs.io API and to AI agents over MCP.