Machine-readable product documentation

Feedback4.dev REST API v1

A workspace- and project-bounded JSON API for projects, reviewers, feedback workflow, approvals, service credentials and signed webhooks.

Feedback4.dev agent interfacehttps://feedback4.dev/api/mcp
REST
/openapi.json
MCP
/api/mcp
Context
/llms-full.txt

01

Executable contract

OpenAPI 3.1 is published at https://feedback4.dev/openapi.json. The original versioned URL, https://feedback4.dev/api/v1/openapi.json, remains available for compatibility.

Every business operation uses Authorization: Bearer f4_live_…. API keys expire after 1–365 days, carry explicit scopes, and may be restricted to one project. The secret is returned once and only its hash is stored.

02

Capabilities

  • Inspect the current key boundary with GET /me.
  • Create and list projects; read widget installation data; configure the project agent mode.
  • List tickets, fetch exact selected-element context, update status/priority/assignee, read the proposed prompt, and request client approval.
  • Authorize or revoke project reviewers and invalidate their active sessions.
  • Manage workspace members and invitations with workspace-scoped keys.
  • Issue or revoke project-scoped MCP credentials and configure durable signed webhook endpoints.

03

First request

curl -fsS https://feedback4.dev/api/v1/me \
  -H "Authorization: Bearer $FEEDBACK4_API_KEY" \
  -H "X-Request-Id: integration-check-001"

04

Operational rules

  • Authenticated responses use Cache-Control: no-store and carry X-Request-Id.
  • JSON write bodies are limited to 64 KiB and reject unknown fields.
  • Collection cursors are opaque. Read limits are 300/minute and write limits are 120/minute per key; 429 responses include Retry-After.
  • A project-scoped key cannot create projects or manage workspace members, even if those scopes were requested.
Give an agent the exact contract, not a marketing guess.

Use the raw Markdown document or OpenAPI specification as context, and issue the smallest project-scoped credential required for the task.

Open the AI-agent runbook