Q for developers

Q Developer Hub

Choose a practical integration path for retrieval, grounded answers, durable agents and extensions.

Build against Q's public /v1 API for retrieval, grounded answers and agent execution. Start with Build with Q to connect your own backend without cloning this repository. For credential-free exploration, use the local sandbox. For a browser product, use the React backend-proxy starter.

The default API root is https://api.qdivision.ai/v1. A deployment's /capabilities and /openapi.public.json are authoritative for its operations, scopes, feature gates and schemas. This checkout includes locally qualified, unpublished SDK 1.1.0 archives; use the matching API and support matrix.

Choose a path

GoalGuide
Complete backend integrationsTested copilot, knowledge and tool-agent recipes · OpenAI client profile
Diagnose a failed or slow integrationTroubleshooting
Prepare to serve real usersIntegration release checklist
Python or TypeScript SDKSDK first answer
First successful integrationBuild with Q from your own application
Understand local fixture limitsSynthetic sandbox
Add Q to a browser productBrowser proxy contract · runnable React example
Retrieve evidence and connect private dataRetrieval permissions · embedding compatibility · index delivery
Add tools, skills and connectorsExtension contracts · host examples
Compose knowledge, memory and documents on a backendHeadless capabilities · executable backend
Tune retrieval in the operator interfaceRetrieval quality — Settings → Retrieval
Consume final answers and streamsPublic schemas and errors · machine answer streaming · evidence policies and availability · verification and structured output
Start and recover agent requestsDurable request and receipt guidance · idempotency · approvals
Verify artifact and webhook deliveryCompletion receipts, downloads and replay
Select public operationsCapability registry · interactive reference
Authorise tenants and usersAuthentication · workspace resources · outbound policy
Upgrade an integrationSDK support and migration · versioning
Develop or release QTest setup · release checks · quality/performance evidence · quality evaluation and rollout · critical-path profiling · fair capacity and overload · cache and routing policy · spending and reconciliation

Start safely

Credentials come from your provisioned Q environment and belong on your backend. Grant only the scopes required by the deployment's advertised operations. A q_test_ prefix does not by itself isolate data or prove that provider calls are free. Use the local synthetic sandbox for credential-free protocol work; see authentication.

For answers, inspect the final sources, citations, evidence and quality. Draft tokens and an open connection are not a successful final result. For agents, retain the operation keys and returned receipt, request Temporal explicitly when required, and poll the original run. Consult operation metadata before retrying a mutation; a new key can create new work.

The examples index distinguishes the qualified entry examples from older, specialised demonstrations. The Q2.5 evidence records clean-environment and browser qualification. Local protocol tests do not certify live provider accuracy, deployment capacity or process-restart recovery.

Managed agent consumers should read the event journal and replay contract, including revision replacement, cursor expiry and result-access errors.

Data use, retention and erasure: Data governance.

Operational release and recovery procedures: Deployment and data recovery.

On this page