The intent layer for the whole company

Your AI agents know the code.
They don't know the company.

Your decisions live in meetings, chat and tickets. Align turns them into one record of what still stands, what conflicts, and why. Your coding agents check it before they edit, and every other agent in the company can too.

Works with any agent that speaks MCP Self-host it, or we host it securely

Under a twentieth of the cost per right answer vs repo search, on decisions never written in the code. Our own benchmark

An illustration: a coding agent, a salesperson and a support agent each check Align before they act, get back the decision that opposes their plan, and change course.

Connects where decisions happen

  • Slack
  • Microsoft Teams
  • Zoom
  • GitHub
  • Jira
  • Confluence
  • Notion
  • Linear
  • Azure DevOps
  • Microsoft 365

When the decision was never written in the code

97%
of agent runs followed it, with Align

Our own benchmark, human-labelled: 6 of its 29 tasks, 5 runs each (30 per arm).

From the meeting to the work that ships

Big company calls and small ones settled in chat, followed across teams to the code.

  1. Capture it where it's decided

    Tag @align in chat, type /align in a comment, or react with an emoji. Align pulls out the decision and why it was made. You confirm it. It can't recover a reason nobody wrote down.

  2. Link it into the Decision Graph

    Each decision links to the intent it serves and the decisions it replaces. When an agent asks later, it gets the decision that still stands, not the newest message about it.

  3. Check it before anything ships

    Your coding agent gets the rule before it writes: from a pre-edit hook where its runtime supports one, otherwise by asking over MCP. Any other agent, or a person about to promise something, can ask the same graph first. Make the Align Decision Check a required status on pull requests and a conflict blocks the merge.

Up and running in minutes. Connect over OAuth and keep the triggers your team already uses.

Your decision graph is yours

  • Open-source core

    The CLI and connector SDK are MIT-licensed. Read the code, run it yourself.

  • Decisions, not messages

    Align only reads conversations you invite it into. It keeps the decision and a source link, and clears the message text after analysis. Imports are held until 33 days after the job last changes. Details

  • You choose where it runs

    Self-host it and Align Tech Ltd has no access. Hosted, it is stored in eu-west-2.

Start on your own, or with your team

Try it yourself

Run align and your graph lives on your machine. Agent edits can be checked against it.

curl -fsSL https://align.tech/install.sh | sh

Open source under MIT, no account, all integrations included. Beta, pre-1.0.

Your team's graph

Team plans start with a demo. We connect your tools, scan your history, and walk you through what it found.

Self-host the whole of Align

Run it in your own Kubernetes cluster with our Helm chart. You bring Postgres with pgvector, your own LLM provider and OIDC sign-in. Align Tech Ltd has no access to your graph.

Growth or Enterprise. One-VM pilot in preview. No Kubernetes? We'd build Nomad or ECS job specs with you.

Align icon align.tech