Documentation Index
Fetch the complete documentation index at: https://docs.atlasmeets.com/llms.txt
Use this file to discover all available pages before exploring further.
The boundary
The gateway is the customer-owned bridge between Atlas and the customer’s private systems.- expose actions
- keep secrets and business logic on the customer side
- translate Atlas requests into whatever internal systems the customer already uses
Why Atlas uses a gateway
Without a gateway, Atlas would need direct access to customer secrets and internal systems. That is the wrong boundary. The gateway keeps the contract clean:- Atlas asks for a capability
- the gateway decides how to execute it
- the customer keeps control
What can sit behind it
- internal REST APIs
- webhooks
- workflow engines
- internal agents
- MCP servers
- OpenClaw-style capability layers
- data stores and search layers
What Atlas sees
Atlas does not need to know what runtime is behind the gateway. It only sees:- a manifest
- an invoke endpoint
- job resources
- optional callbacks
One workspace gateway
Today, the clean model is one gateway per workspace. That gateway can expose many actions. Atlas agents then get different subsets of those actions.Good action design
Keep actions task-shaped. Good:fetch_internal_contexttrigger_followup_workflowlookup_candidate_packetquery_research_agent
- raw model completion endpoints
- thin wrappers around low-level internal APIs with no clear task boundary
The v1 contract
Gateway Contract v1 is built around:GET /v1/manifestPOST /v1/invokeGET /v1/jobs/{jobId}POST /v1/jobs/{jobId}/cancel
- async callbacks
- approval flows
What it is not
Gateway actions are not Atlas skills.- Atlas skills live on the Atlas side and shape behavior
- gateway actions live on the customer side and perform work