1 /

Every era needed a new tool.
The AI era is no different.

SCCS
1972
SCCS
Single-file deltas
CVS
1986
CVS
Multi-file commits
SVN
2000
Subversion
Atomic commits
Git
2005
Git
Distributed branching
h5i
now
h5i
shared versioned context for AI agents

The code survives.
The reasoning doesn't.

"Why did Claude touch src/session.rs?
That wasn't in the ticket."

— Every engineering team, every week

Git records

  • Which lines changed
  • Author and timestamp

Git misses

  • Why those files were touched
  • What the AI read before editing
  • Where the AI was uncertain

A Git sidecar that versions
the context behind AI work

Stores structured metadata in .git/.h5i/ — no new infrastructure, no lock-in, works with any git remote.

h5i capture commit --prompt …

AI provenance

Prompt, model, agent ID, and test results linked to every commit.

h5i recall notes

Session analysis

Files read vs. edited, uncertainty heatmap, blind-edit detection.

h5i recall context

Reasoning workspace

Version-controlled OBSERVE / THINK / ACT trace that survives session resets.

h5i audit compliance

Compliance & policy

Enforce AI commit rules and generate audit reports across date ranges.

Every agent inherits
what the last agent learned

h5i codex prelude

Session prelude

A new agent starts with the branch goal, milestones, open TODOs, and the last useful decisions.

h5i recall context show

Reasoning workspace

OBSERVE, THINK, ACT, NOTE, branch, merge, and checkpoint entries form a Git-native context DAG.

h5i share pr post

PR as one surface

When the branch reaches review, h5i renders the same context into a sticky GitHub PR comment.

~/my-project
$ h5i codex prelude

[h5i] Context workspace active
goal: ship OAuth login safely
milestones: provider integration done;
            token refresh in progress
TODO: integration test for failover path

Last decisions:
  THINK Redis survives process restarts
  ACT   switched session store to Redis

$ h5i share pr post --style review
# same context, rendered for reviewers

Reasoning that survives
session resets

Version-controlled OBSERVE / THINK / ACT trace stored in refs/h5i/context. Snapshots on every commit. Branches like git.

h5i recall context trace

Trace entries

Log each reasoning step as OBSERVE, THINK, or ACT — persists across session resets.

h5i recall context search "…"

Targeted retrieval

BM25 over traces + git co-change. Replaces exploratory Read chains with ranked candidates.

init · feat/auth
OBSERVE auth.rs ×4
THINK use Arc injection
branch: experiment/mutex
THINK try Mutex
✗ abandoned
ACT inject Arc into middleware
commit · "analyzed auth"
OBSERVE tests/ ×2
commit · "updated tests"

6× fewer file reads.
Same answer.

Agents burn tokens every session re-deriving what they already figured out last time. h5i capture claim records each conclusion with its evidence pinned as a Merkle hash over the files it depends on — stays live until any evidence blob changes, then auto-invalidates.

h5i capture claim "…" --path src/api.rs

Content-addressed facts

Evidence = sha256 over (path, blob_oid) pairs at HEAD. Git tells you when they're stale — no TTL, no guessing.

h5i recall context prompt

Preamble injection

Live claims render as ## Known facts at session start. Agent treats them as pre-verified — skips re-reading.

A/B evaluation
── A/B · 5 trials per arm ─────────────────
  metric            No claims   With claims      Δ
  ────────────────  ─────────   ───────────  ─────
  Read tool calls   6.2 ± 0.4     4.0 ± 0    −35%
  Cache-read tok.    793,425       511,824    −35%
  Assistant turns  25.6 ± 3.8   16.6 ± 3.1    −35%
  Est. session cost    ~$4.35        ~$2.13    −51%
  Task fidelity           5/5          5/5  ✓

  ✔  All 5 seeded trials read exactly the 4
     relevant HTTP files.

Enforce rules.
Generate audit reports.

h5i audit policy init

Policy rules

Define TOML rules — require audit flag, minimum prompt length, model allowlist, test coverage threshold.

h5i audit policy check

Pre-commit enforcement

Checks staged files against policy before every commit. Blocks on errors, warns on violations.

h5i audit compliance --since …

Audit report

AI contribution ratio, audit coverage, policy violations, and top-risk files over any date range.

~/my-project
$ h5i audit policy check

   model declared (claude-sonnet-4-6)
   audit flag present
   test metrics missing          [error]
   prompt length < 20 chars      [warn]

  1 error · 1 warning — commit blocked

$ h5i audit compliance --since 2026-04-01

── Compliance Report ──────────────────────
  AI-generated:    74% of commits (30 days)
  Audit coverage:  91%
  Policy errors:   3  warnings: 8
  Top risk file:   src/billing/token.rs

Everything you need to
audit AI-generated code

h5i capture commit --audit

Integrity audit

12 deterministic rules — credential leaks, CI tampering, scope creep.

h5i recall blame <file>

AI-aware blame

Per-line authorship with agent name, model, and test badge.

h5i rewind <sha>

Safe rewind

Non-destructive restore to any past commit. WIP auto-backed up.

h5i audit scan

Injection detection

Scans the reasoning trace for prompt-injection patterns.

h5i capture memory

Memory tracking

Snapshots Claude's memory at every commit, diffs across versions.

h5i serve

Web dashboard

Timeline, Integrity, Intent Graph, Memory, Sessions at localhost:7150.

h5i vs. plain Git

Capability
Plain Git
h5i
Who changed what
AI provenance — prompt, model, agent
Which files the AI read, not just edited
Blind-edit and uncertainty detection
Reasoning persists across sessions
Targeted retrieval — no exploratory reads
Works with existing git remotes