For agents
MakerPerks is built to be read by machines as well as people. Everything below is generated from the same source of truth as the site, so it never drifts. Public, read-only, no key required.
Query it with MCP
The official Model Context Protocol endpoint lets your agent query the directory through read-only tools — ask "what can an X claim?" instead of fetching and filtering the whole dataset.
Connect (Claude Code / Claude Desktop)
claude mcp add --transport http makerperks https://mcp.makerperks.com/mcp
Add it at user scope (-s user) to make it available across
all your projects. Other MCP clients: point a Streamable-HTTP transport
at
https://mcp.makerperks.com/mcp (legacy SSE clients can use
/sse).
Tools
-
search_perks(query?, audience?, category?, min_value?) - Search programs by text, persona, category, or minimum value.
-
perks_for_persona(persona) - Every perk a persona qualifies for, highest value first.
-
get_perk(slug) - One full program record by slug.
-
list_personas() - The personas, with per-persona counts.
-
list_categories() - The category tags, with frequencies.
Read-only and rate-limited — no accounts, no writes, no tracking.
Results come from the same data as perks.json, so they
reflect the latest deploy.
Raw data
Prefer to fetch the dataset directly? Three formats, same source:
- perks.json
- The full machine-readable dataset — every program with value, audience, tags, and provenance.
- llms.txt
- A curated index grouped by persona, following the llmstxt.org convention.
- llms-full.txt
- The same index with every program inlined, for a single-fetch read.
Structured metadata
Every program detail page embeds schema.org JSON-LD, so crawlers and assistants get typed offer data inline — no separate fetch needed.