Keep your crew. Watch it work.

Sulcus instruments your crew’s normal kickoff calls. CrewAI still owns the crews, agents, tasks, tools, delegation and output.

Hosted by Sulcus, CrewAI 1.14+

Sulcus adds

Visibility across the kickoff.

  • Crew, agent and task activity that CrewAI exposes.
  • Model and tool events from CrewAI’s hooks and event bus.
  • Reported token usage and an optional per-run token limit.
  • Stop control and a bounded event history for each run.

CrewAI keeps

The crew and its work.

  • Crew and task definitions, processes, delegation and planning.
  • Agent selection, prompting and tool invocation.
  • The crew’s result and anything your app stores.
  • Side effects your application or providers have already started.

How it attaches

Around your normal kickoff calls.

When a project selects CrewAI, Sulcus activates its instrumentation in the run’s container and wraps kickoff and akickoff. Your application kicks off its crew as usual.

Observation boundary
Runs the workflow
CrewAI
Instrumented calls
Crew kickoff / akickoff
Event source
CrewAI hooks and event bus
Not collected by instrumentation
Prompts, model outputs, tool payloads

What this integration doesn’t do.

The workflow tree is shallower than for LangGraph, because CrewAI exposes less structure. Sulcus doesn’t create agents, assign tasks or choose delegation paths. Sulcus supports CrewAI 1.14 and later 1.x releases.

Sulcus doesn’t offer tool permission settings or approvals for hosted crews. Stopping a run and the token limit can’t undo a provider request or side effect that has already started. What your app prints is kept with the run, so avoid printing sensitive data.

See your first run in Sulcus.