Kleio

How it works

Capture decisions where they happen.
Query them when you need them.

Kleio drops into your existing workflow via MCP, CLI, and GitHub App. No new tools to learn. No process to enforce.

Step 1

Capture

Decisions, checkpoints, and work items flow in from MCP (in-editor), CLI, and GitHub App. Both human and AI decisions are captured automatically.

Step 2

Connect

Signals are linked across sources. Backlog items are synthesized with semantic deduplication and Eisenhower prioritization.

Step 3

Query

Ask kleio_ask any question about your engineering history. Get answers grounded in actual decisions, not hallucinated summaries.

The capture layer

Three capture surfaces, zero behavioral change.

MCP Server (in-editor)

Capture decisions and checkpoints directly from Cursor, Claude Code, or VS Code. The MCP server runs alongside your agent and captures context without interrupting your flow.

Works with: Cursor, Claude Code, VS Code, any MCP-compatible editor

CLI

Capture decisions, checkpoints, and work items from your terminal. Integrates into existing scripts, hooks, and automation. Works where MCP doesn't reach.

Commands: kleio capture, kleio decide, kleio checkpoint, kleio check

GitHub App

Automatically captures PRs, reviews, commits, and CI signals. Builds provenance trails that connect decisions to the code changes that implemented them.

Captures: PRs, reviews, commits, CI/CD status, branch events

The signal pipeline

Captured signals become connected, structured, actionable.

Signal linking

Decisions, checkpoints, PRs, and commits are connected through signal links. When a PR implements a decision, the link is automatic. When a checkpoint references a backlog item, they're connected.

Backlog synthesis

Work-item captures are synthesized into structured backlog items with Eisenhower prioritization (urgency + importance). Semantic deduplication prevents the same issue from being filed twice.

Auto-close

High-confidence semantic matches from checkpoints or merged PRs can auto-close open backlog items. Explicit KL-N references in commit messages also trigger resolution.

Provenance trails

Every decision, checkpoint, and backlog item links back to the PRs, commits, and contributors involved. Trace any line of code back to the decision that created it.

The query layer

Ask why. Get real answers grounded in your team's decision history.

kleio_ask (RAG)

Natural language queries over your entire engineering history. Answers are retrieved from actual decisions, checkpoints, and signals — not generated from thin air.

Backlog views

Filter, search, and triage your synthesized backlog. See what follow-up work is pending, who captured it, and what it connects to. Prioritize with confidence.

See it in action

An agent plans a migration. Kleio captures every signal along the way.

Agent Plan

Task: Migrate auth from Firebase to Clerk

  • 1.Scaffold Clerk provider and session hooks
  • 2.Replace Firebase login/signup flows
  • 3.Update API middleware for Clerk JWT validation

Press Build to run the capture demo.

Your AI agents already make decisions. Kleio makes sure they're remembered. Try it free →

Ready to try Kleio?

Start free, connect your workspace, and capture decisions from day one.