Quickstarts by capture path
BlackLake observes AI actions through eight capture paths. Each path is a different way work reaches the control plane — they all converge on the same policy, cost, and receipt model. Pick the one that matches how your team operates.
mcpMCP gateway
Govern Claude Code, Codex, Cursor, and any other MCP-compatible CLI through BlackLake. The fastest path for coding tools.
sdkSDK govern()
Wrap any LLM call in your application code with `bl.govern({...})`. Policy decisions, cost capture, signed receipts — all from a single call.
ciCI pipeline
Gate deploys, code-mod runs, and other CI actions on a governed BlackLake decision. Same SDK; the actor is the pipeline.
shellShell wrapper (blx)
`blx <command...>` — gate any shell command on a governed decision before running. The capture path for terminal-driven AI agents.
cloud-auditCloud audit ingest
Forward GCP, AWS, Azure, and GitHub audit events to `/v1/audit/ingest`. Coverage for actions taken via the cloud console or out-of-band tools.
existing-workflow-engineExisting workflow engines
Adapters for Temporal, Inngest, Trigger.dev, and BullMQ. Same govern() call, lineage columns populated by the engine.
depthDepth durable workflows
BlackLake's own durable workflow runtime — `depth run <workflow.ts>`. Workflow lineage flows into the receipt automatically.
manualManual / console-driven
Operators create agents, tools, and policies directly in the console. No code change required.