Agent-readyPrepaid — no subscriptionReal field data

Research API

Deep research over 400+ real, outcome-labeled tech support resolutions — actual completed jobs, each ending with the fix that worked, written by the technician who was there. One POST returns an evidence-grounded brief your product or agent can act on.

✓ Payment received — your API key is on its way to your email (only copy, keep it safe).

Quickstart

curl -X POST https://api.bloomtechsupport.com/api/v1/research/corpus \
  -H "X-API-Key: YOUR_KEY" -H "Content-Type: application/json" \
  -d '{"problem": "HP OfficeJet says paper jam but there is no jam"}'

The response is a research brief, not a chat completion:

{
  "summary": "Seen repeatedly in this corpus: phantom jam errors are
              usually a paper-size mismatch, not a mechanical jam...",
  "corpus_had_relevant_cases": true,
  "cases": [ { "title": "...", "snippet": "...worker-confirmed fix..." } ],
  "proven_steps": [ "Check tray paper-size setting vs loaded paper", ... ],
  "pitfalls": [ "Do not start with a rear-access teardown..." ],
  "clarifying_questions": [ ... ],
  "credits_remaining": 24
}

One successful run = one credit. Balance any time: GET /research/credits with the same header. Runs take 20–60 seconds — the agent is genuinely reading case files, not doing a vector lookup.

Why it answers differently than a frontier model

Frontier models guess from general knowledge. This API researches what actually fixed similar jobs — every case file ends with a worker-confirmed resolution from a real service call, de-identified before storage. Independently scored by our Last Mile Exam: on about 1 in 3 problems the corpus contains the exact working fix, over half for recurring job types like computer repair. About the underlying dataset →

Credit packs — prepaid, from $0.40 per run

Checkout takes a minute; your key arrives by email and never expires. No subscription, no card on file. For comparison: Gemini Deep Research runs ~$2–5 per task — wiring this into a job workflow (estimate call + mid-job call) costs about $1 per job on the $50 pack.

$20 · 10 runs · try it

Buy

$50 · 100 runs · 50¢/run

Buy

$100 · 250 runs · 40¢/run

Buy

Building a product on this, or need higher rate limits / other verticals? contact@encoders.ai

For AI agents

This page and /llms.txt are agent-readable. Keys are scoped to research only. Limits: 6 requests/minute per key, platform-wide concurrency cap (429 + Retry-After when busy), 402 when credits run out. Errors are plain JSON.

Async mode (recommended — runs outlive many client timeouts): POST /research/corpus/jobs with the same body returns 202 + a job id; poll GET /research/corpus/jobs/{id} until status is done (the brief is in brief). Same credits, same key.

OpenAPI: openapi.json — generate a client or hand it to your agent.

MCP server — mount the corpus as a tool in Claude Code or any MCP client (Streamable HTTP):

claude mcp add encoders-research https://api.bloomtechsupport.com/api/v1/research/mcp \
  --transport http --header "X-API-Key: YOUR_KEY"

One tool: research_past_resolutions. Each call burns one credit, same as REST.