Control what agents do.
See everything.
Platform policies define what agents are allowed to do. Agent profiles standardize configurations across teams. Full audit trails track every session, tool call, and cost.
Platform Policies
SAM platform policies define rules, constraints, and delegation boundaries for agent behavior. Policies are injected into agent instructions so agents operate within the boundaries your team sets. Policies propagate to child tasks automatically — when one agent dispatches work to another, the child inherits the parent's policy set.
Agent Profiles
Engineering leads create profiles that bundle an agent type, model, and configuration into a reusable template. Teams pick from approved profiles instead of configuring from scratch. New team members get a working setup immediately.
When standards change (new model, different agent, updated system prompt), update the profile. Every future task uses the new configuration. No per-user migration needed.
Session Audit Trails
Agent sessions are recorded: what was requested, what the agent produced, and task lifecycle events (start, completion, failure). Project-scoped activity feeds give team leads visibility without requiring access to the agent's workspace. When using SAM-managed AI billing, token usage and cost are tracked per session.
Comments and Credential Attribution
Reviewers comment directly on an agent's message or a library file, and the thread becomes part of the record — sent to the agent as context, then marked resolved. Nothing gets lost to a side channel like chat or email.
On shared projects, a Credential Attribution panel shows which triggers, tasks, nodes, and deployments still run on an individual's personal API key or cloud credential versus a shared project credential, with a Fix link to move them over.
Admin Observability
The admin dashboard provides real-time and historical views of platform health:
- Worker log streaming and error tracking
- VM agent health and heartbeat monitoring
- Cost monitoring with projections, by model, day, and user
- AI usage breakdowns by model and user
- Per-user compute usage and node-hour quota tracking
- Node and workspace lifecycle tracking
Knowledge & Context Management
SAM's knowledge graph stores project context, user preferences, and organizational decisions that persist across agent sessions. Agents start every task with the context they need instead of rediscovering it from scratch. Knowledge is project-scoped and searchable.
Get your coding agents off your laptop
SAM is open source. Deploy it to your Cloudflare account and connect your agent. Add a cloud provider when you want full environments on your own servers.