Search 48.4M knowledge entries or bring your own data for indexing. Tokenized with a 2.8M whole-word vocabulary. Zero storage on our end.
New here? Read why deterministic retrieval matters or what Deterministic Knowledge Infrastructure is.
https://cold-api.coldstate.aiAll authenticated endpoints require a Bearer token in the Authorization header:
Authorization: Bearer cs_live_...The short version is below. The full quickstart also covers building and searching your own index, and the MCP guide lists every tool.
Create a free account and copy your API key (cs_live_…). No credit card required. Quotas and rate limits for each plan are on the pricing page.
Search the global knowledge base — no index setup needed. Drop in your key and run:
curl -X POST https://cold-api.coldstate.ai/v1/search/global \
-H "Authorization: Bearer cs_live_..." \
-H "Content-Type: application/json" \
-d '{"query": "deterministic retrieval", "limit": 5}'Prefer MCP? The same API key works as a hosted MCP server — nothing to install. New to MCP? Read What is an MCP server? Otherwise, add this to your client's MCP config and restart it:
{
"mcpServers": {
"coldstate": {
"url": "https://cold-api.coldstate.ai/mcp",
"headers": { "Authorization": "Bearer cs_live_..." }
}
}
}Claude Code CLI: claude mcp add --transport http coldstate https://cold-api.coldstate.ai/mcp --header "Authorization: Bearer cs_live_..."
/v1/search/global/v1/search/domains/v1/knowledge/fetch/v1/knowledge/cite/v1/knowledge/verify/v1/knowledge/snapshot/v1/knowledge/related/v1/knowledge/resolve/v1/knowledge/batch/v1/knowledge/isomorph/v1/knowledge/stats/v1/knowledge/capabilities/v1/indexes/v1/indexes/v1/indexes/:id/v1/indexes/:id/documents/v1/indexes/:id/download/v1/indexes/:id/v1/index/v1/indexes/:id/search/v1/search/v1/indexes/:id/explain/v1/auth/register/v1/auth/login/v1/auth/forgot-password/v1/auth/reset-password/v1/auth/keys/v1/auth/keys/create/v1/auth/keys/rotate/v1/auth/keys/:keyId/deactivate/v1/auth/keys/:keyId/v1/auth/logout/v1/usage/v1/usage/history/v1/billing/subscription/v1/billing/portal/v1/billing/cancel/v1/billing/resume/mcp/openapi.json/llms.txt/.well-known/mcp/server.json/health/readyErrors are returned as JSON with the HTTP status repeated in the body:
{ "error": "...", "message": "...", "statusCode": 429 }Queries are limited to 2000 characters, batches to 10 queries, and results to 100 per query. Quotas and rate limits for each plan are on the pricing page.
Every search response includes a routing state that describes the topology of the results:
Build an index on our infrastructure, download the .db file, and run searches locally. Zero storage on our end — we wipe the file after download or after the 24-hour TTL.
POST /v1/indexes with mode: "iaas" and your documents. Status: queued → processingGET /v1/indexes/:id until status is ready. Response includes download_url and download_expires_at.GET /v1/indexes/:id/download to stream the .db file. Server copy is deleted immediately. Status: deliveredqueued → processing → ready (24h download window) → delivered / expiredEvery result carries a primary relevance score plus two diagnostic signals. All three are deterministic — identical for identical inputs — so you can cache, compare, and audit them.