Azure Foundry Services — Agent Tools via Responses API
OpenClaw is the agent orchestrator. Foundry provides managed AI services that OpenClaw agents can use via the Responses API. You do NOT create Foundry agents — you call Foundry tools directly.
Endpoint
All requests: http://localhost:8443 with ?api-version=2025-11-15-preview. Auth is automatic.
Calling Foundry Tools via Responses API
Code Interpreter (run Python)
curl -s -X POST 'http://localhost:8443/openai/responses?api-version=2025-11-15-preview' \
-H 'Content-Type: application/json' \
-d '{"model":"gpt-4.1","input":"Calculate fibonacci(20) using Python","tools":[{"type":"code_interpreter","container":{"type":"auto"}}],"store":false}'
Web Search (Bing grounding)
curl -s -X POST 'http://localhost:8443/openai/responses?api-version=2025-11-15-preview' \
-H 'Content-Type: application/json' \
-d '{"model":"gpt-4.1","input":"What are the latest Rust language features?","tools":[{"type":"bing_grounding"}],"store":false}'
Memory Search (cross-session recall)
curl -s -X POST 'http://localhost:8443/openai/responses?api-version=2025-11-15-preview' \
-H 'Content-Type: application/json' \
-d '{"model":"gpt-4.1","input":"What beverage does the user prefer?","tools":[{"type":"memory_search","memory_store_name":"kars-memory","scope":"default"}],"store":false}'
Querying Foundry Infrastructure
List existing agents in the project
curl -s 'http://localhost:8443/agents?api-version=2025-11-15-preview'
List model deployments
curl -s 'http://localhost:8443/deployments?api-version=2025-11-15-preview'
When to use
- Call Foundry tools (code interpreter, web search, memory) via Responses API
- Inspect what Foundry agents or models exist in the project
- Use Foundry as a service layer for OpenClaw agent capabilities
When NOT to use
- Do NOT create Foundry agents — OpenClaw is the agent orchestrator
- For memory CRUD operations (use foundry-memory skill)
- For simple chat completions (use /v1/chat/completions directly)