Technical documentation for developers
The public integration contract for developers defines the Python SDK as the main application interface and four platform behaviors outside the task loop.
Technical documentation for developers covering the Python SDK, agent harness, Libraries, web search, model routing, and the documentation MCP server.
- Python SDK is the main application interface for submitting typed tasks and hosting tools.
- Platform runs the agent loop, reads Libraries, searches the web, and picks model routes.
- Documentation MCP server lets supported AI clients search and cite these pages directly.
- Covers errors, reliability, security, scope and limits, content marking, and agents and access.
Summaries were generated by AI. Generative AI is experimental.
Duale AI’s technical documentation covers the public integration contract for developers. The Python SDK is the main application interface.
Four platform behaviors run outside your application’s task loop. The platform runs the agent loop with its built-in tools. It reads documents from the Libraries you grant. It can search and read the public web. For each attempt, it picks a model route from the pool your tenant enabled. The documentation MCP server lets supported AI clients search and cite these pages directly.
When a page directs you to the Dashboard, use the URL from your deployment handoff. The managed service uses the Duale AI Dashboard.
Start here
- Python SDK—install the SDK, provision access, submit typed tasks, host tools, and attach documents.
- Errors and reliability—diagnose a task that does not finish, typed SDK failures, document ingestion failures, retries, and timeouts.
- Agent harness—learn what you stop building, what your agent can reach without you, and what you still own.
- Libraries—give your agents documents to read, and keep the boundary around what each one reaches.
- Web search—learn how an agent finds and reads public pages, and what source access does not prove.
- Model routing strategy—design tenant model pools, regional boundaries, endpoint redundancy, and model upgrades. It also carries the deployment handoff a network or platform architect needs: endpoints, private connectivity, and the paths each environment has to reach.
- Agents and access—divide agent ownership between teams, and see which boundaries stop at the tenant.
- Security—review the shared responsibility split, data protection, the threat model, and the evidence available for an audit.
- Scope and limits—review intended use, operational limits, records, and customer duties.
- Content marking—check which answers carry a machine-readable mark and how to verify one offline.
- Documentation MCP server—connect these docs to Codex, Claude, ChatGPT, Cursor, or another MCP client.