Research API for AI coding agents.
Give Codex, Claude Code, Cursor, or your own agent a fast research tool for dependency choices, migration risk, market scans, and PR-ready evidence. It returns citations, typed JSON, and hard budget limits instead of another open-ended browser loop.
- ~90s
- Lite mode for quick agent checks
- 20
- sources in Standard mode
- 202
- async response for background work
“Can I safely migrate this Vite dashboard to React Router v7 in this PR?”
A coding agent asks for fresh docs, changelogs, and repo context before editing.
Fluxweave searches, fetches, chunks, and filters sources under a fixed budget.
The agent receives a recommendation, citations, risk notes, and next steps for the PR.
curl https://api.infrence.ai/v1/research \
-H 'Authorization: Bearer inf_live_…' \
-H 'Content-Type: application/json' \
-d '{
"question": "Find the current React Router v7 breaking changes before editing this routing PR",
"mode": "lite"
}'{
"id": "9b0a4f12-7e3a-4c2b-9e1d-3f4c5a6b7c8d",
"status": "succeeded",
"credits_charged": 20,
"brief": "React Router v7 keeps most v6 route objects intact, but framework mode and data APIs changed around bundling, typegen, and server/runtime config. For this PR, keep the current BrowserRouter path unless you also plan the framework-mode migration.",
"typed": {
"recommendation": "stay_on_library_mode_for_this_pr",
"risk": "medium",
"agent_next_steps": [
"scan for createBrowserRouter usage",
"avoid framework-mode file routes in this diff",
"open a follow-up migration issue"
]
},
"sources": [
{ "url": "https://reactrouter.com/upgrading/v6", "title": "Upgrade from v6", "domain": "reactrouter.com" },
{ "url": "https://reactrouter.com/start/framework/routing", "title": "Framework routing", "domain": "reactrouter.com" },
{ "url": "https://github.com/remix-run/react-router", "title": "React Router repository", "domain": "github.com" }
],
"latency_ms": 72184
}Pick a latency envelope per agent task.
AI coding agents should not wait for deep research on every turn. Route quick checks to Lite, deeper migrations to Standard or Pro, and background investigations to Max.
| Agent task | Mode | Wall-clock | Credits | Output |
|---|---|---|---|---|
| Pre-edit dependency check | Lite | ~90s | 20 | 5 sources, fast brief |
| PR planning or migration research | Standard | ~4 min | 65 | 20 sources, synthesis pass |
| Risky architecture decision | Pro | ~8 min | 245 | 60 sources, citation check |
| Background deep dive | Max | up to 15 min | capped | 120 sources, metered |
Research your agent can safely use.
Fluxweave turns the parts coding agents struggle with into explicit primitives: source policy, cost ceilings, structured answers, and auditable citations.
Use Lite for loop-time dependency checks, Standard for PR planning, Pro when the agent needs citation validation.
Pass a JSON Schema and get structured fields like recommendation, risk, sources, and next_steps.
Each answer returns title, URL, domain, and source summaries so humans can audit the agent before merge.
Return 202 immediately, stream progress over SSE, or post a signed webhook when the research is ready.
Allowlist docs, changelogs, GitHub, package registries, or block domains your agent should never cite.
Every request accepts a credit ceiling. Failed validation refunds, and cancelled jobs release reserved credits.
Cost controls for unattended agents.
$0.01 per credit on packs. No subscription required, no card on file, and every request can set the maximum spend before work begins.
Save when agents run research every day.
Heavy usage? Subscribe for lower per-credit cost and higher concurrency.
Same endpoint, different depth.
Agents can choose cheaper, faster research for small code edits and reserve deeper work for decisions that will survive code review.
| Mode | Credits | List price | Includes |
|---|---|---|---|
| Lite | 20 | $0.20 | 5 sources, ~90s |
| Standard | 65 | $0.65 | 20 sources, ~4 min |
| Pro | 245 | $2.45 | 60 sources, ~8 min, citation check |
| Max | metered | your max_credits | Up to your max_credits, 120 sources, ~15 min |
Inspect real research before wiring it into your agent.
Browse published briefs or submit a topic. It is the fastest way to judge citation quality, answer shape, and the kinds of questions Fluxweave handles well.
Put fresh research in your coding agent loop.
500 free credits, no card required. Start with a Lite request, then move deeper when the agent needs more evidence.