Agents and workflows
Agents are the specialists; workflows put them on a schedule. Both run inside the same governance.
The agents hub
The hub carries a large library of ready-made agents across sales, finance, engineering, HR, operations, legal, and more. Each agent brings a role, a toolset, and working instructions. Pin the ones your team relies on so they are one tap away in the composer.
Build your own
- Open the agent builder and describe the role in plain language: what the agent is for, how it should work, and what it must never do.
- Choose its tools. Only the tools you grant are available, and destructive tools stay approval-gated no matter what the instructions say.
- Save it to your organisation. Versions are kept, so you can refine instructions over time without losing what worked.
Renly can also form a specialist agent for you mid-task: describe the job and the composer proposes an agent with a scoped toolset and an execution passport that records exactly what it is allowed to do.
Workflows
- Multi-step runs: chain steps with conditions, retries, and per-step timeouts.
- Scheduling: run workflows on a schedule in your timezone.
- Approval gates: mark a step as requiring approval and the run pauses until a human decides. Bulk-approve when a batch is ready.
- Run history: every run keeps its steps, outputs, and token usage. Retry from a failed step instead of starting over.
Follow-ups and background work
Agents can schedule their own follow-ups: ask for a check-in tomorrow morning and the agent re-runs server-side at that time and posts its report into the same chat. Longer jobs can spawn sub-agents that work in parallel and report back into one place.