Real workflows teams run on Nairi
Four playbooks we use ourselves and ship as docs: on-call, PR review, knowledge Q&A, ad-hoc data. Each one is a real agent setup with the rules, MCPs, and safety constraints that make it work in production.
SRE / platform / on-call
On-call incident response
An agent in the incident channel that pulls metrics, tails logs, correlates against recent deploys, drafts status updates, and co-authors postmortems. Read-only by default; write actions opt-in per tool. Self-hostable.
See the playbookEngineering / dev tools
Automated PR review
A GitHub Action calls the Nairi REST API on PR open. The agent runs your tests and linter in its sandbox, posts inline review comments to GitHub, and summarizes back in Slack. Per-repo skills and rules, harness choice per repo.
See the playbookOps / people / any team with a wiki
Company brain on Slack
A shared agent connected to Notion, Confluence, or Drive. Answers "how do we do X here?" with citations. Can file Linear tickets, post in other channels, or open PRs to fix the docs when they're wrong. Confidentiality dial built in.
See the playbookData / analytics / non-SQL teammates
Data analyst on Slack
Wire your warehouse (Postgres, Snowflake, BigQuery) as a custom MCP. Anyone on the team asks plain-English questions; the agent writes the SQL, runs it, and posts the answer with the table. Three layers of safety on every query.
See the playbookThe four above are the playbooks we've documented
Equip an agent with MCPs and skills, point it at any internal API - what gets built is up to you. And many more beyond these four.
Spin up an agent that fits one of these
Connect the tools, encode the rules, mention the agent. Each playbook ships with a working docs setup and a demo video.