Inputs
- Deal events emitted across the workspace
- Per-agent state and outstanding approvals
Orchestration Layer
Keeps the agents working from the same context.
The orchestration layer is what makes the stack behave like one team instead of a dozen disconnected tools. It carries context between agents, sequences work across the deal, and routes decisions that need your approval to you rather than acting unilaterally.
Orchestration LayerIn production
01
Coordinated hand-offs between agents
02
A daily digest of what moved and what needs you
03
Approval requests for anything consequential
Last reviewed 2026-07-20
What it works from, what it produces
Concrete inputs. Concrete outputs.
No abstraction — this is the actual work the agent performs inside your workspace.
Outputs
- Coordinated hand-offs between agents
- A daily digest of what moved and what needs you
- Approval requests for anything consequential
Limits
What it will not do. Stated up front.
An agent you can trust is one whose boundaries you know before you buy.
Consequential actions surface for your approval rather than executing silently
Included with the Autopilot tier, which is capped while provisioning is productized
Evidence
What we can actually show.
Every claim on this page names where it comes from and when it was last checked.
Source: docs/agent-catalog.md, orchestrator section · as of 2026-07-04
Shipped and running in production with per-customer orchestration and a daily operator digest.
The rest of the team
It does not work alone.
Each agent below has its own evidence page. The others on the roster appear on the team overview until their delivery evidence supports a page of their own.
See it on your workflow
Watch Orchestration Layer run against your own pipeline.
A walkthrough uses your numbers and your process — not a canned demo script.