Serving the HTTP API

meshflow serve starts a FastAPI server exposing REST, SSE, and WebSocket endpoints.

Start

meshflow serve \
  --host 0.0.0.0 \
  --port 8000 \
  --db meshflow_runs.db \
  --policy-file policies/production.yaml \
  --otlp-endpoint http://localhost:4318

For development with colored output:

meshflow dev

Key endpoints

Method	Path	Description
GET	/health/live	Liveness probe — 200 if process alive
GET	/health/ready	Readiness probe — 200 if ready to serve
GET	/events	SSE stream of all workflow events
GET	/ws/bus	WebSocket message bus
GET	/runs	List recent run IDs
GET	/runs/{run_id}	Get all steps for a run
POST	/runs/{run_id}/approve	Approve a paused HITL checkpoint
GET	/eval-results	List stored eval results
GET	/plugins	Discovered plugins
GET	/analytics	Workflow analytics (last N runs)
GET	/otel/config	Live OTEL exporter status
GET	/compliance/report	Generate compliance report
GET	/graph/{run_id}	Workflow graph for a run
GET	/sla	SLA stats across all agents

CLI client

meshflow logs              # recent run history
meshflow replay <run_id>   # step-through debugger
meshflow approve <run_id>  # approve HITL checkpoint

Python client

from meshflow import MeshFlowClient

client = MeshFlowClient("http://localhost:8000", api_key="mf-...")
result = client.run_agent("assistant", "What is 2 + 2?")
print(result.output)

# SSE streaming
for event in client.stream_events():
    print(event.event_type, event.data)

API keys

meshflow keys generate --name prod-key --role operator
# → mf-xxxxxxxxxxxxxxxx

# Use in client
MeshFlowClient("http://...", api_key="mf-xxxxxxxxxxxxxxxx")
# Or via header: Authorization: Bearer mf-...

Environment variables

Variable	Description
MESHFLOW_HOST	Default host (overrides --host)
MESHFLOW_PORT	Default port (overrides --port)
MESHFLOW_DB_PATH	Ledger database path
MESHFLOW_POLICY_FILE	Policy YAML path
MESHFLOW_OTLP_ENDPOINT	OTLP span export endpoint