ProjectDNA
LIVE BETA · 3 PILOT SLOTS OPEN
Engineering memory for AI coding agents

Every commit has a story.
Your AI should know it.

ProjectDNA reads every PR, commit and ADR in your repo and gives Claude, Cursor and Copilot the institutional memory your team already has — so your AI ships cheaper, more accurate, and stops rewriting code because it forgot the why.

Start your pilot →Watch 60s demo ▶
Works with◇ Claude↗ Cursor⌘ Copilot~ Windsurf// Cline
−60%tokens·0hallucinations·10 minsetupCLAUDE · CURSOR · COPILOT · WINDSURF · CLINE
−60%
Less tokens per session

Compressed memory means agents stop re-reading your repo every time. Real customers, real measurements.

0 hallucinations
Every answer cites its source

PRs, commits, ADRs — the agent can’t make up a function name when the answer is grounded.

Less rework
Better code, fewer regressions

Agents remember why a decision was made — they stop reinventing and breaking it.

10 min setup
Read-only GitHub App

No source code stored. Signed DPA. Works with the AI tools your team already uses.

Inside ProjectDNA

A queryable library of your project — not a copy of your code.

For every PR, commit, branch, ADR and doc, we generate a compressed memory document with sources. Your AI retrieves what it needs — never trains on your code.

  • PR memoryWhat changed, why, files affected, risks introduced
  • Commit memoryCoalesced summaries — trivial commits filtered automatically
  • Branch memoryCurrent goal, progress, open questions, suggested next steps
  • ADR memoryOne memory per decision document, kept fresh on edit
  • Doc memoryREADME + docs/ + architecture/ — what your team has already written
  • Project memoryThe whole repo, distilled — purpose, stack, domains, instructions
PR#42 · auth: refresh-token rotation2d
merged · a8f3c1 · cited 12× by agents
ADR07 · Async memory generation1w
accepted · docs/adr/07.md
COMMIT9c2e · refactor billing module3d
feat/billing · 4 files · 217 lines
DOCdocs/api/payment.md3d
last touched · cited 5× by agents
BRANCHfeat/auth · 12 commits aheadlive
open PR #103 · waiting for review
PR#88 · BullMQ for memory pipeline1w
merged · b71f · cited 8× by agents
ADR12 · Tenant isolation strategy2w
superseded by ADR-14
COMMITe3a0 · README updates5d
main · README.md
How it works

Three steps. Ten minutes. One pilot.

Install once. Memory generation runs continuously in the background. Your agents start querying within minutes.

01

Connect

Install the read-only GitHub App on the repos you want. No source code ever stored. Webhook signature verification, encrypted tokens, signed DPA.

~30 seconds
02

Remember

We summarize every PR, commit, branch and ADR into compressed memory — automatically, in real time. Secret scanning + PII redaction before persistence.

~5 minutes for first repo
03

Answer

Your AI agents query ProjectDNA via the MCP server. Cited answers (PR #, commit hash, ADR title). Sub-second. Branch-aware. Sourced.

unlimited queries
Security & privacy

Your repo. Your data.
Always.

ProjectDNA was built by engineers who'd never connect this to their own repos if it weren't safe to.

READ FULL FAQ →

No source code stored

We fetch diffs on demand, hold them encrypted for ≤24h, then drop them. We store only summaries.

No AI training on your data

Signed DPAs with Anthropic (ZDR addendum) and OpenAI (Enterprise). Monitored monthly.

Tenant isolation, verified

A 100-pair cross-org fuzz test runs on every code change we ship. Cross-tenant leaks fail the build.

GDPR + LGPD compatible

Signed DPAs, Standard Contractual Clauses, tamper-evident audit logs. Export anytime. Hard delete within 30 days. Regional residency available on Enterprise.

Pilot program — 3 slots open

Be one of the first three teams
to give your AI real memory.

Pilot includes: up to 10 repos, all 6 memory types, audit logs, founder-priority support, DPA + sub-processor agreement, manual invoicing. €79/mo for the duration of the pilot.

NO CREDIT CARD · CANCEL ANYTIME · DPA INCLUDED