Agent integration

Configure workload credentials

Register an AI Agent workload in the dashboard, then keep its API key in your secret manager. New clients authenticate with headers rather than putting credentials in the URL.

export KAVROS_API_URL="https://<data-plane-host>/api/agent/egress"
export KAVROS_API_KEY="<workload-api-key>"
export KAVROS_AGENT_ID="<workload-id>"

Node.js CLI

Use the dependency-free @kavrosai/cli for smoke tests, scripts, jobs, or any workflow that does not need a native SDK.

npm install --global @kavrosai/cli
kavros doctor

kavros request \
  --action post_data \
  --target "https://api.example.com/v1/run" \
  --content '{"input":"hello"}' \
  --header 'content-type:application/json' \
  --json

For one-off use, replace the global install with npx @kavrosai/cli doctor.

Native HTTP from any language

The same JSON contract works from Go, Rust, Java, Ruby, TypeScript, or any HTTP client. Send the workload credentials as headers:

curl --fail-with-body --silent --show-error \
  --request POST "$KAVROS_API_URL" \
  --header "Authorization: Bearer $KAVROS_API_KEY" \
  --header "X-Kavros-Agent-ID: $KAVROS_AGENT_ID" \
  --header "Content-Type: application/json" \
  --data '{
    "action": "post_data",
    "target": "https://api.example.com/v1/run",
    "content": "{\"input\":\"hello\"}",
    "headers": {"content-type": "application/json"}
  }'

Policy blocks return HTTP 403. Older clients may use query parameters for compatibility, but new integrations should use headers.

Python zero-code hook

The optional kavros-agent wheel patches requests and httpx at startup, covering Anthropic/OpenAI SDKs and plain HTTP clients without application code changes.

python -m pip install kavros-agent
python app.py

The hook is a no-op unless the Kavros environment variables are set. See the native HTTP path above for other runtimes.