# Pecta

## Docs

- [GET and POST /v1/agents — manage agents](https://docs.pecta.ai/api-reference/agents.md): List all agents in your organization or register one by ID. Agents are created automatically on first evaluation — explicit registration is optional.
- [Authenticate your requests to the Pecta API](https://docs.pecta.ai/api-reference/authentication.md): Learn how to use publishable and secret API keys, set the Authorization header, and sign telemetry batches with HMAC-SHA256 for the Pecta API.
- [POST /v1/evaluate — cloud gate evaluation](https://docs.pecta.ai/api-reference/evaluate.md): Submit an agent output to Pecta's cloud gate engine. Get an instant pass/fail verdict, per-gate results, and an updated reputation score in one response.
- [GET /v1/events — list evaluation events](https://docs.pecta.ai/api-reference/events.md): Retrieve a paginated list of evaluation events for your organization. Filter by agent, gate verdict, or time range to find exactly the events you need.
- [POST /v1/ingest — batch telemetry ingestion](https://docs.pecta.ai/api-reference/ingest.md): Ingest up to 100 evaluation events in one batch. The @pecta/core SDK calls this endpoint automatically — most customers never need to call it directly.
- [Pecta REST API: endpoints, auth, and errors](https://docs.pecta.ai/api-reference/overview.md): Complete reference for the Pecta REST API. Authenticate with a Bearer token, call endpoints under https://api.pecta.ai, and receive JSON responses.
- [GET /v1/reputation/:agentId — fetch agent score](https://docs.pecta.ai/api-reference/reputation.md): Fetch the current reputation score and lifecycle status for an agent. Scores reflect a rolling window of the last 500 evaluations and update in seconds.
- [Quality gates in Pecta: how they run and fail](https://docs.pecta.ai/concepts/gates.md): Quality gates are parallel, fail-fast pass/fail checks on agent output that run in-process with no network overhead and no training period needed.
- [How Pecta agent reputation scores are computed](https://docs.pecta.ai/concepts/reputation.md): Every agent_id gets a portable 0–1000 score computed from pass rate, latency, streaks, and volume — updated over a rolling window of 500 evaluations.
- [How Pecta handles privacy and API key security](https://docs.pecta.ai/concepts/security.md): Pecta never stores payloads or user content. Gates evaluate locally; only pass/fail metadata reaches the cloud. Learn how API keys and HMAC signing work.
- [Monitor agent reputation scores in the dashboard](https://docs.pecta.ai/dashboard/agents.md): Track every agent's reputation score, pass rate, and gate failure history. Drill into per-agent detail views to diagnose quality issues over time.
- [Browse agent evaluation events in real time](https://docs.pecta.ai/dashboard/events.md): Browse every agent evaluation result with filters by agent ID and outcome. Page through the full log or narrow to failures in seconds.
- [Pecta dashboard: real-time agent monitoring](https://docs.pecta.ai/dashboard/overview.md): Monitor AI agent quality at a glance. The Pecta dashboard shows live evaluation events, gate analytics, and aggregate pass rates for your org.
- [Content gate: reject empty or refused responses](https://docs.pecta.ai/gates/content.md): Block agent outputs that are empty, whitespace-only, or contain phrases where the model declines to answer instead of completing the task.
- [Build custom quality gates for Pecta engines](https://docs.pecta.ai/gates/custom.md): Implement the Gate interface to add domain-specific pass/fail checks to any engine, alongside or instead of Pecta's built-in gates.
- [Filesystem gate: block dangerous shell commands](https://docs.pecta.ai/gates/filesystem.md): Detect destructive rm commands, path traversal sequences, and references to sensitive system directories anywhere in agent output before execution.
- [Latency gate: enforce response time limits](https://docs.pecta.ai/gates/latency.md): Fail any agent output whose upstream latency exceeds a configurable millisecond threshold, with automatic skip when no latency is recorded.
- [Pecta built-in gates: complete reference](https://docs.pecta.ai/gates/overview.md): Every gate Pecta ships out of the box, how to compose them in createEngine, and how the timeout budget and fail-fast behavior work together.
- [PII gate: detect personal information in output](https://docs.pecta.ai/gates/pii.md): Scan every string in agent output for email addresses, Social Security Number patterns, and US phone numbers before the response leaves your system.
- [RTB adomain gate: reject placeholder advertiser domains](https://docs.pecta.ai/gates/rtb/adomain-verify.md): Detects placeholder, test, and malformed advertiser domains in bid responses, ensuring every bid carries a legitimate and properly formatted adomain.
- [RTB audience safety gate: COPPA compliance checks](https://docs.pecta.ai/gates/rtb/audience-safety.md): Enforces COPPA compliance by detecting child-directed inventory signals and blocking bids whose creative categories fall into an unsafe content type list.
- [RTB bcat gate: enforce blocked content categories](https://docs.pecta.ai/gates/rtb/bcat-compliance.md): Fails any bid whose creative categories overlap with the request's bcat blocked list, enforcing the OpenRTB blocked-category contract on every response.
- [RTB bid sanity gate: catch unrealistic bid prices](https://docs.pecta.ai/gates/rtb/bid-sanity.md): Flags bids whose price is unrealistically high relative to the impression floor, catching stale configs, decimal-place bugs, and currency mismatches.
- [RTB impid gate: validate bid impression IDs](https://docs.pecta.ai/gates/rtb/impid-match.md): Verifies that every bid.impid in the response corresponds to a valid imp.id in the originating OpenRTB request, catching phantom and mismatched bids.
- [RTB gates: OpenRTB quality checks for bidders](https://docs.pecta.ai/gates/rtb/overview.md): Six in-process gates that validate OpenRTB bid responses against their originating requests — designed for DSPs, SSPs, and ad exchanges.
- [RTB tmax guard: skip stale bid evaluations](https://docs.pecta.ai/gates/rtb/tmax-guard.md): Skips all downstream gates when the OpenRTB tmax deadline budget is nearly exhausted, preventing wasted work on bids that are already too late.
- [Schema gate: validate agent output with Zod](https://docs.pecta.ai/gates/schema.md): Enforce a strict structural contract on agent output or input using any Zod schema, with per-issue failure details returned for debugging.
- [How Pecta works: modes, gates, and reputation](https://docs.pecta.ai/introduction.md): Learn how Pecta's three integration modes — SDK, MCP proxy, and REST API — evaluate agent output with quality gates and build portable reputation scores.
- [Evaluate agent output via the Pecta REST API](https://docs.pecta.ai/quickstart/api.md): Send agent output to POST /v1/evaluate from any language and get a gate verdict, evaluation ID, and live reputation score back in a single HTTP response.
- [Gate MCP tool responses with pecta-proxy CLI](https://docs.pecta.ai/quickstart/proxy.md): Gate MCP tool responses without touching server code. pecta-proxy runs locally with no key or with a publishable key for cloud telemetry and reputation.
- [Get started with the @pecta/core Node.js SDK](https://docs.pecta.ai/quickstart/sdk.md): Install @pecta/core, create an engine with gates, evaluate your first agent output, and ship telemetry to the Pecta cloud — all in under 15 ms.

## OpenAPI Specs

- [openapi](https://docs.pecta.ai/api-reference/openapi.json)