Skip to main content
The coding agent is what actually does the work. Agents run when you start a session manually or when an automation triggers them. Key characteristics:
  • Adaptive: Agents decide the best path, not a fixed script.
  • Context‑aware: They use your repo, tickets, and integrations.
  • Collaborative: Progress and results show up in Slack and the web.

​
How it works

When a session starts, Proliferate:
  1. Spins up an isolated environment (your repo, Docker, dependencies)
  2. Injects the agent with context (trigger event, instructions, available actions)
  3. Agent runs autonomously — reads code, makes changes, runs tests
  4. Agent outputs results (PR, preview, message, etc.)

​
What agents can access

The agent has access to:
  • Full filesystem — your repo
  • Terminal — run any command
  • Actions — GitHub, Slack, Sentry, etc. (based on permissions)
  • Context sources — pull info from integrations