Data & API

AgentSpot publishes its full directory as static JSON at stable, versioned URLs. No key, no rate limit, CORS enabled — built for agents and developers.

Start here

These are files, not a query service: there are no ?q=, ?tag=, or ?sort= parameters, and adding one is ignored. Search is served by shape instead — fetch the compact index, filter it in your own process, then fetch the single listing you care about. The index is a fraction of the size of the full dataset and carries every field you need to shortlist.

curl -s https://agentspot.dev/data/index.json     # all 81, search fields only
curl -s https://agentspot.dev/data/listings/agent/openclaw.json   # one listing, every field

Endpoints

URL	Contents
`/data/index.json`	Compact index. Every active listing (81) with search fields only. Fetch this first.
`/data/listings/<id>.json`	A single listing, in full — the `id` verbatim, e.g. `/data/listings/agent/openclaw.json`.
`/data/all.json`	Every active listing with every published field.
`/data/agents.json`	Agents (27), in full.
`/data/skills.json`	Skills (15), in full.
`/data/tools.json`	Tools (21), in full.
`/data/coding.json`	Coding (10), in full.
`/data/learn.json`	Learn (8), in full.
`/data/schema.json`	JSON Schema (2020-12) for every payload above.
`/data/openapi.json`	OpenAPI 3.1 description of these endpoints.
`/.well-known/agent.json`	Discovery document — every URL on this page, machine-readable, so nothing has to be guessed.

Shape

Every payload is the same envelope: a _license attribution block, a generated timestamp, an api_version, a count, and a listings array — even the single-listing files, so one parser handles all of them. Internal bookkeeping (raw metric history, review notes) is never published.

{
  "_license": "CC-BY-4.0",
  "_license_url": "https://creativecommons.org/licenses/by/4.0/",
  "_attribution": "Data from AgentSpot (https://agentspot.dev)",
  "api_version": 2,
  "generated": "…",
  "count": 81,
  "listings": [ … ]
}

Listing fields

Fields marked index also appear in the compact index; the rest are in the full payloads only.

Field	Type	Meaning
`id` index	`string`	Stable identifier, `<category>/<slug>`. Never reused.
`name` index	`string`	The project's own name.
`category` index	`string`	One of: agent, skill, tool, coding, education.
`description` index	`string`	A one- or two-sentence summary written by us — never copied from the README.
`url` index	`string`	Canonical AgentSpot page for this listing.
`repo_url` index	`string`	Source repository. Read this before running anything.
`website_url`	`string ∣ null`	The project's own site or docs, when it has one.
`license`	`string`	SPDX identifier where GitHub could classify it, otherwise "Other (see repo)".
`package` index	`object ∣ null`	Registry coordinates for the published package, when one exists. `null` means the project is installed from source.
`install` index	`string ∣ null`	One-line install command derived from `package`. A starting point, not a substitute for the project's own instructions — check the repo for required flags, runtimes, and configuration.
`can_run_locally` index	`boolean ∣ null`	Whether the project can run entirely on your own machine, with no required hosted service. `null` where the question does not apply (a book, a prompt pack) or has not been assessed.
`is_mcp_server` index	`boolean`	Heuristic: the project describes itself with MCP (Model Context Protocol) topics. A shortlisting hint, not a verified capability — we do not currently publish per-listing MCP endpoint URLs or config snippets.
`open_source`	`boolean`	False for listings whose source is not publicly available.
`first_party`	`boolean`	True for tools published by the lab or vendor whose model they serve.
`derivative_of`	`object ∣ null`	Set when this listing is a fork that has overtaken the project it came from; points back at the original.
`github_stars` index	`integer ∣ null`	Stars at `last_verified`.
`downloads`	`object ∣ null`	Downloads over the trailing 7 days, from the registry named in `source`. Container pulls count restarts as well as installs, so they are discounted in the score.
`popularity_score` index	`number ∣ null`	Composite 0–100 score. `null` when the listing is `unranked`. The formula is published at /methodology/.
`unranked`	`boolean`	Listed and shown, but deliberately excluded from popularity ordering. `editorial_note` says why.
`growth`	`object ∣ null`	Change over the trailing week, held steady between weekly updates. `null` until a listing has enough history.
`tags` index	`array`	The project's own repository topics, lightly cleaned.
`first_listed`	`string`	The date this listing first appeared here.
`last_verified`	`string`	The date the metrics above were last refreshed from public APIs.
`editorial_note`	`string ∣ null`	A public note from us, where a listing needs one.
`sponsored`	`object`	Sponsorship never affects inclusion, ordering, or `popularity_score`. `active` is false on every listing today.
`tier`	`string`	"standard" for every listing today; reserved for a future curated tier.
`curation`	`object ∣ null`	Reserved for future independent evaluation results. `null` today.

On install: it is derived mechanically from the package a project publishes, and it is a starting point rather than an instruction. Nothing here is vetted; read the repository, pin the version, and prefer rebuilding a capability from source over installing an artifact blindly. See the disclaimer.

Stability

api_version is 2. Listing ids are stable and never reused. New fields may be added without a version bump — parse defensively and ignore what you do not recognise; removing or retyping a field would bump the version.

Licensing & attribution

Factual data (names, star counts, links) is not copyrightable, but we publish the compiled dataset under CC-BY-4.0: reuse it freely with attribution to AgentSpot (https://agentspot.dev). The same license is declared in every payload and in llms.txt.

For LLMs & answer engines

An orientation file for AI crawlers lives at llms.txt (mirrored at /.well-known/llms.txt), with a full prose index of every listing at llms-full.txt. Every page carries Schema.org JSON-LD, and new listings are announced in the RSS feed. We deliberately welcome AI crawlers — see our robots.txt.