Augmentcode Client Agent Rollout (Onboarding)

Augmentcode Client Agent Rollout (Onboarding)

1. 01

   Confirm the client's integration surface before provisioning

   Verify the client has GitHub, Slack, Jira, and PagerDuty accounts your team can authorize. Cosmos coordinates agents across those four systems, so a missing PagerDuty connection means incident triage cannot run and the retainer scope should exclude it.

2. 02

   Provision the Business plan seat allocation

   The Business plan is $100/month for up to 50 seats. Map seats to named client engineers, not agency staff, so the client owns the license at contract end and your delivery margin sits in the setup fee rather than a resold seat markup.

3. 03

   Connect repositories and set the Context Engine scope

   Point Cosmos at the client's primary repository only for the pilot. The Context Engine maps codebases structurally and cuts token consumption by 33% versus keyword-based retrieval, but scoping to one repo keeps the first billing cycle predictable.

4. 04

   Enable the code review expert on pull requests

   Turn on the pre-built code review expert against the pilot repo and set human-in-the-loop policy so agent comments post as suggestions until the client's tech lead signs off on two weeks of output.

5. 05

   Wire the ticket-to-PR and testing experts

   Connect Jira so the ticket-to-PR expert converts assigned tickets into reviewed pull requests, and attach the test coverage expert to one core application module. Both are pre-built experts, so configuration is policy and routing, not custom build.

6. 06

   Add incident triage through PagerDuty

   Route PagerDuty alerts to the incident management expert so agents investigate and summarize before a human is paged. Agree an escalation threshold with the client in writing, since triage output feeds directly into their on-call rotation.

7. 07

   Integrate CLI access into the client CI/CD chain

   Install CLI access into the client's existing pipeline so agents run inside their build tooling rather than a parallel system. This is the step most likely to need client DevOps time, so book it as a joint session.

8. 08

   Hand off the runbook and close the retainer loop

   Document every agent configuration, policy, and escalation threshold in a runbook the client's dev team owns. Schedule a 30-day review to confirm throughput gains against the 2-3x claim before quoting any expansion into migrations or security remediation.