Skip to content

Imbi

Imbi is a DevOps service management platform for large environments with many services and applications.

What is Imbi?

  • A service catalog. Imbi records each project, its type, its owner team, its links, its dependencies, and its documents, in a graph.
  • A delivery record. Imbi records the commits, pull requests, tags, releases, and deployments of each project, and shows which environments run different code.
  • A quality check. Imbi scores each project, checks it for problems (the Doctor), and records the components that each release ships.
  • An automation platform. Imbi runs AI agents, workflows, and scheduled tasks against your projects, and connects to the tools you already use through plugins and webhooks.

Key features

Catalog

Delivery

  • Releases and deployments: deploy, promote a release to the next environment, block a release, and stop a deploy of a commit with failed CI
  • Drift: see where environments run different code, and which commits are not released
  • Pull requests and dependencies: pull request and commit history, and the software components of each release (SBOM)
  • The data model is in Delivery

Quality

AI and automation

  • Agents and agent tasks: AI agents that do work against your projects, in runner pools that you deploy
  • Workflows: deploy, rollback, agent, email, decision, and wait steps that start manually, on a schedule, or from a webhook
  • Prompts, AI models, and secrets: versioned prompts, the model catalog of each organization, and short-lived secrets for agents
  • Scheduler: cron, interval, and one-shot triggers that call the API or the gateway
  • MCP server and Slack bot: use Imbi from MCP clients and from Slack

Platform

Architecture

The services are in one container image:

Service Purpose
imbi-api The REST API, authentication, authorization, and the background workers
imbi-assistant The AI assistant in the UI
imbi-gateway Receives webhooks and runs plugin actions
imbi-mcp The MCP server for AI agents and MCP clients
imbi-scheduler Fires scheduled tasks against imbi-api and imbi-gateway
imbi-slackbot The Slack bot
imbi-workflows The workflow engine
Caddy and the UI The reverse proxy, and the static files of the React UI

Caddy is the only public port. imbi-slackbot connects out to Slack in socket mode, and imbi-workflows has no public route.

These parts are not in the image:

  • imbi-harness, the agent runner, has its own image and Helm chart. See Agent Runners.
  • iggy-connect runs from the Apache Iggy image as a separate container. It writes the analytics rows to ClickHouse.

See the architecture overview.

Data stores

Store Required Purpose
PostgreSQL + Apache AGE Yes The graph of the catalog, and the relational schemas of agent tasks, workflows, secrets, runner pools, and the scheduler
ClickHouse Yes Analytics: the operations log, events, commits, pull requests, scores, drift verdicts, runs, and audit trails
Apache Iggy Yes Message streams. Every ClickHouse row goes through Iggy, and iggy-connect writes it to ClickHouse
S3-compatible storage Yes Uploaded files
Valkey For background work The work queues of the imbi-api background workers (syncs, scoring, maintenance)
SMTP server For email Welcome and password reset email, and workflow email steps. Email is on by default