AgentWire
A read-only API over a small, cited incident record
AgentWire is a read-only HTTP API and Atom feed over a curated record of AI-agent incidents. The dataset is deliberately small — 4 records today — and every entry cites its original source. Read it as a working shell for a data product, not as an established intelligence feed.
What it is
AgentWire reads a ledger of incident records and republishes them. It does not collect, enrich, scrape, or contact any external system. Each record carries a citation to its original source, and summaries are reproduced as recorded.
| Endpoint | Returns |
|---|---|
| GET /health | Service status and a dataset summary, including the record count. |
| GET /v1/incidents | A page of records, newest first, filterable by severity, type, and chain, with limit and offset. |
| GET /v1/incidents/:id | One record by its content-addressed id. |
| GET /v1/digest | Counts by type and severity, the date range, and the most recent entries. |
| GET /v1/feed.xml | An Atom 1.0 feed of the record, newest first. |
JSON responses use one error shape, and only GET and HEAD are accepted. The full reference is in products/agentwire/docs/API.md.
Who it is for
AI-security teams
Keep a citable record of agent-related incidents next to your own detection work.
GRC and risk
Pull incident categories and severities into reporting without maintaining the record yourself.
Underwriters and analysts
Consume the feed programmatically as one input among several.
Data provenance
Stated plainly
- AgentWire does not collect data. It republishes records that already exist in the ledger.
- The dataset is small. At the time of the current build it contains four records. The tooling is complete; the record is what grows over time.
- Each record is a third-party report or primary-source item. The source field carries the citation.
- Records are not verified findings and not advice. Some are allegations or hazard scenarios rather than adjudicated harm.
- Severity is a derived field used for filtering and display, not a ledger field. The rule is documented alongside the data model.
Pricing
| Plan | Price | Scope |
|---|---|---|
| Feed | $99 / month | Read access to the Atom feed and the list endpoint, with citations. |
| API | $499 / month | Full endpoint access, filters, pagination, and the digest for programmatic use. |
Pricing reflects access to the service and the record as it stands. It is not a subscription to a specific incident count or a forecast of coverage.
Limits
Keep these in view
- AgentWire is not advice of any kind, and its records are not verified findings.
- The dataset is small and does not represent the field. Absence of an incident in the record is not evidence that it did not happen.
- Coverage depends on what has been recorded and cited; it is not exhaustive and not a statistical sample.
- The service is read-only and makes no outbound requests. It re-publishes citations that remain attributable to their original sources.