Align
Product Self-hosting Pricing Docs Contact Book a demo
Free and open source

Install the CLI

It is open source under MIT, and there is no account to create. One binary, no Node required. Install it, point it at the tools you already use, and any MCP agent can read your decision graph.

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

Prefer npm? npm i -g @aligndottech/cli

View on npm Beta, pre-1.0

Want the hosted version for your team?

That is Growth: one shared graph across everyone's tools and agents, rather than your own. There is no self-serve signup, so we set each team up ourselves. Book a demo, or leave your email for updates.

Or skip the wait and book a call.

Sources
  1. 1 Qodo, 2025 State of AI Code Quality (n=609): 65% of developers using AI to refactor report that it misses relevant context (around 60% for writing, testing and reviewing).
  2. 2 Dillon & Varanasi, Context-Augmented Code Generation (arXiv:2605.08112): augmenting Claude Code with a product-context system - recorded decisions, personas, customer signals, competitive intelligence, spec generation and mid-build consultation - took decision compliance from a 46% baseline to 95% across 8 tasks and 41 weighted decision points. A third-party benchmark of a different product, not Align data; the paper does not isolate decision retrieval's individual contribution from the rest of that system.
  3. 3 Sonar, 2026 State of Code Developer Survey (n=1,149, fieldwork October 2025): 96% do not fully trust AI output; 48% always verify it before committing.
  4. 4 CodeRabbit, State of AI vs Human Code Generation, December 2025 (n=470 PRs: 320 AI-co-authored, 150 human-only): AI-co-authored PRs carry about 1.7x more issues, counted in a code-review taxonomy rather than as production defects.
  5. 5 Weller, Lawrie and Van Durme, NevIR: Negation in Neural Information Retrieval (arXiv:2305.07614): IR models are asked to rank two documents differing only by a negation. The authors report that most models, including state-of-the-art ones, perform the same as or worse than a random ranking, and that a large gap to human performance remains.
  6. 6 monday.com engineering, I Am Morphex, 16 April 2026: their account of running production coding agents on their own codebase, including the two mistakes they name, one of which is over-investing in vector stores.
  7. 7 Align's own measurement: a contradicting decision pair scored 0.31 by cosine similarity against an agreeing pair at 0.38. Our data, on our own graph, and the reason we adjudicate rather than rank.
  8. 8 Cortex, Engineering in the Age of AI: 2026 Benchmark Report (50+ engineering leaders): incidents per pull request up 23.5% year on year.
NVIDIA Inception Program member © 2026 Align Tech Ltd. Company registered in England & Wales No. 16719662.
Install the CLI • Why not build this? • Whitepaper • Proof • Docs • Privacy Policy • Terms of Service
Align icon align.tech

We value your privacy

We use cookies to see how people use the site. We never sell your data.

By continuing to use our site, you agree to our Privacy Policy.