Skip to main content

Welcome to Align

Align is the decision graph for engineering teams. Your agents know the code. They don't know the company.

ADRs catch the big calls. The hundreds of smaller ones - most decided in meetings, the rest scattered across chat, tickets and PRs - leave both your engineers and your agents blind to what still stands, what conflicts, and why. Align makes them one graph you can ask.

The shape of it, in one pass​

The order matters, and it is the thing most people get wrong on first read. A commit is an artefact of a decision, not a record of one. So the tools split into two roles:

CAPTURE                          THE GRAPH                    VALIDATE
where decisions are made         what Align builds            where code meets the decision

Slack · Teams                    decisions, evidence,         GitHub
Jira · Confluence      ──────▶   participants, and     ──────▶  ·  a PR is checked against
meeting transcripts              11 relationship types           what was already decided
                                 between them
                                        │
                                        │  MCP server / CLI / API
                                        ▼
                      People · AI coding agents · Dashboards · Audit

GitHub is on the right of that diagram. Align does not read your commits to work out what you decided; it checks your pull requests against what you already decided somewhere else. If you only remember one thing from this page, remember which side of the arrow GitHub is on.

What you actually get​

Ask the graph. Your agent, your CLI, or you: "why is auth built this way?", "what did we decide about the payments retry policy?", "what happens when the alignment check can't run?" One answer - what still stands, who decided it, and a link to where. Not seven Slack threads and a stale Confluence page.

This is the primary use. No model of ours sits between the agent and the record: the MCP tools return the stored decision, not a generated answer.

Get warned before you contradict one. Every alternative - a RAG bot, a CLAUDE.md, enterprise search, your own vector store - waits to be queried. The Agent Alignment Layer surfaces the conflict at the point of action, whether or not anyone thought to ask. And when it cannot run, it says unknown rather than quietly passing.

Onboard against the why, not the what. A new joiner reads the code to learn what was built. The graph is where they learn why, and which of the three attempts at it is the one that survived.

Capture is always an explicit act​

Align does not sit in your channels listening. Nothing is captured until a human does something deliberate:

TriggerWhere
@align <decision>Slack, Teams
/alignJira and Confluence comments, GitHub PRs and issues
React with 📌 or 🔖Slack, Teams (configurable)
Post-meeting transcriptTeams, Zoom
Discover - a historical scanopt-in, per source, reviewed before anything is saved

Ambient suggestions are off by default and opt in per channel, project, repo or space, and nothing is stored until someone confirms it. See Capture for the full matrix of what each platform can and cannot do.

Start where you are​

⌨️ CLI

Free and open source (MIT). Build a personal decision graph from the tools you already use, then expose it to any AI agent over MCP. No account required.

CLI Quickstart →

☁️ Pro (Cloud)

One shared decision graph across the whole org, captured from every team's tools, with cross-team conflict detection. Hosted by Align.

Get Started →

🏠 Self-Hosted

Run the entire stack inside your own infrastructure. Data sovereignty, SSO, air-gap support.

Install Guide →

See the full plan comparison. No integration is ever paywalled individually, on any plan, and there is no cap on how many you connect.

Key features​

FeatureDescription
Ask the graphSemantic and full-text search over every decision, from the UI, the CLI, or an agent over MCP
Decision GraphDecisions linked to evidence, participants and one another by 11 relationship types - not an isolated list
Agent Alignment LayerAn agent checks a proposed action before taking it; unknown when the check cannot run, never a silent pass
Conflict & supersession detectionWhich decisions contradict each other, and which one replaced which
Decision SpacesHierarchical org scoping, up to four levels, with cross-space dependency tracking
DiscoverHistorical scan across Jira, GitHub, Slack, Teams, Confluence, Zoom and Linear - reviewed before anything is saved
MCP ServerProvider-agnostic. Claude, Cursor, Copilot, or anything else that speaks MCP

Align complements ADRs, it does not replace them. It carries the everyday decisions that never make it into a formal record, and it links the formal ones to where they were actually argued out.

What Align does not do​

Being clear about this saves everyone a call.

  • It does not reverse-engineer a decision from code. If nobody recorded why, the why is gone. Nobody can recover it. What Align guarantees is that from the day you connect it, the reasoning stops disappearing.
  • It does not write your code. An engineer still has to read the decisions and interpret them. Align makes sure they are reading the right ones.
  • It does not listen passively. Every capture path starts with a human action or an explicit, per-resource opt-in.

Next steps​