SLA.directory API & MCP
One dataset, two ways in: pull SLA data for 300 vendors as plain JSON, or connect the MCP server and ask your AI assistant in plain language. Free, no account.
JSON API
Static JSON, served from a CDN, no authentication, no rate limits.
/api/v1/. Breaking changes ship as /api/v2/.Base URL: https://sla.directory/api/v1/
# every vendor, as a summary list curl https://sla.directory/api/v1/vendors.json # one vendor in full detail curl https://sla.directory/api/v1/vendors/stripe.json
Response envelope
Every response is wrapped in a small envelope. List endpoints carry total; single-resource endpoints carry last_modified. The payload is always under result.
# list response { "schema_version": "1.0.0", "generated_at": "2026-06-20T00:00:00.000Z", "total": 300, "result": [ /* ... */ ] } # single-resource response { "schema_version": "1.0.0", "generated_at": "2026-06-20T00:00:00.000Z", "last_modified": "2026-06-19", "result": { /* ... */ } }
Endpoints
Summary list of all 300 vendors (slug, name, category, uptime, max credit, claim window).
Try it ↗Every vendor in full detail, in a single request (bulk consumers and agents).
Try it ↗Full detail for one vendor: credit policy & tiers, claim process, services, support tiers, exclusions.
Example: stripe.json ↗Vendor summaries within a single category.
Example: finance-and-payments.json ↗Flat search index of all vendors and every individual service (300 vendors plus their services), ideal for client-side search and agents.
Try it ↗services[] array: each entry has name, category, uptime, sla_eligible, sla_url, and a last_verified date (per-service when re-verified individually, otherwise the vendor's). The search.json index lists every service as its own type: "service" entry, so you can resolve a single service straight to its vendor.Notes
- Every endpoint is also available as YAML: swap
.jsonfor.yaml(e.g./api/v1/vendors/stripe.yaml). - Data changes at most daily, so please cache responses client-side and use
If-Modified-Since. - Field names may still change before the
v1stable release (currently beta). - Full schema & field reference: DATA_SCHEMA.md ↗
- Spotted stale data? Suggest a correction.
MCP for AI assistants
Ask about SLAs in plain language. SLA.directory maintains an MCP server ↗ that connects this API to assistants like Claude and Cursor, with answers that cite the source.
Add the MCP server
The server is open source and runs locally over stdio. Build it once from the sla-mcp repo ↗ (Node 18+):
# clone and build
git clone https://github.com/self9dmin/sla-mcp
cd sla-mcp
npm install && npm run buildThen point your assistant's MCP config at the built server, using the absolute path to dist/index.js (Claude Desktop: claude_desktop_config.json; Cursor: mcpServers in settings):
// mcpServers configuration
{
"mcpServers": {
"sla-directory": {
"command": "node",
"args": ["/absolute/path/to/sla-mcp/dist/index.js"]
}
}
}sla-directory-mcp is not on npm yet. Once it is published, the build step goes away and the config becomes "command": "npx" with "args": ["-y", "sla-directory-mcp"]. Until then, use the from-source build above.Example prompts
Once connected, ask questions like these:
How it works
The MCP server is a thin client over the JSON API above. Your assistant calls it to look up vendors, services, credit tiers, and claim processes, then answers in plain language with a link back to the source. No account, no API key, no rate limits.