MarketLens / Intelligence API
Intelligence, with a paper trail.
Discover a research agent. Inspect its evidence. Verify the outcome.
01 / Explore
Try the complete research loop
Create an isolated, one-hour sandbox with linked events, evaluated agents, and successful and failed tasks. All market records are demonstration fixtures. No payment or trading access.
- 01Discover
Find a version by capability.
- 02Inspect
Read its evidence and limits.
- 03Verify
Run a task. Trace the outcome.
Explore the API reference
Intelligence API · v1
From evidence to an inspectable result.
Query the same MarketEvent model used by the human research desk. Select an immutable agent version, authorize its read scope, then inspect the result, trace and capability measurements.
Credentials & boundaries
A sandbox expires after 1 hour. Its key is shown once, stored server-side only as a digest, and kept in memory by this page. Save it privately to reconnect after a reload. Do not put it in URLs, public code or browser storage.
Keys grant access only to their workspace and declared scopes. Sandbox data becomes inaccessible at expiry and eligible for opportunistic cleanup 24 hours later. No live provider or arbitrary remote agent is connected.
Errors & retries
Send Idempotency-Key with every POST. Reuse it when retrying the same requirement or task. Changing the payload returns 409. Executions are version-locked and terminal outcomes are replayed without duplication.
Key creation is a one-time disclosure: repeated creation requests cannot recover a lost secret. List and revoke a lost key, or create a new sandbox. Responses include a request ID for audit inspection.
401: reconnect. 403: missing scope. 404: record not in this workspace. 409: conflicting retry or active task. 422: missing benchmark / incompatible version. 429: rate limited. 503: service unavailable.
Start with a scoped query
curl 'https://marketlensresearch.com/api/v1/network/events?limit=10' \ -H "Authorization: Bearer $MARKETLENS_API_KEY"
API map
| Route under /api/v1/network | Use |
|---|---|
GET /events?q=&asset=&category=&limit=&cursor= | Search snapshots; cursor pages up to 50 |
GET /events/:id | Source → event → assets → reactions → evidence → explanation |
GET /events/:id/{source,evidence,affected-assets,reaction-timeline,mechanisms,explanations,confidence,provenance,analogues} | Typed graph sections; analogues are shared-asset fixture matches |
GET /economic-releases | Demonstration economic releases |
GET /filings and /influential-activity | Unavailable: verified historical providers are not connected |
GET /agents?capability=&verified=true&maxCostUsd=&maxLatencyMs= | Discover version-specific candidates |
POST /agents; POST /agents/:id/versions | Register a built-in adapter and publish immutable configuration |
GET /versions/:id/passport; POST /versions/:id/evaluations | Inspect or run fixture verification |
POST /requirements; POST /tasks; POST /tasks/:id/executions | Create a requirement, lock a version, authorize and execute |
GET /tasks; GET /tasks/:id; GET /benchmarks | Outcomes, traces, expected answers and ledger |
GET /keys; POST /keys; DELETE /keys/:id; GET /audit | Issue narrower keys, revoke access and inspect request IDs |
Request shapes
{
"requirement": {
"eventId": "evt_…",
"capability": "retrieval",
"objective": "Retrieve the recorded headline",
"maxCostUsd": 0,
"maxLatencyMs": 5000
},
"task": {
"requirementId": "req_…",
"agentVersionId": "av_…",
"scopes": [
"intelligence:read"
]
}
}All records preserve demonstration, extracted-fact, derived-inference and authored-explanation distinctions. A confidence score is not proof of causation. Fixture verification is not a claim of trading performance.