oobo | Context beyond the diff | oobo

Context beyond the diff.

Every commit gets a memory. The transcripts, the decisions, the agent runs. A point you can hold onto when the codebase drifts.

curlbrewWindows

curl -fsSL https://oobo.ai/install.sh | bash

macOS · Linux

commit 9f2c84a
Author: agent <cursor@oobo>
Date:   Mon Mar 14 09:41:22

fix auth middleware

src/auth.rs  | 42 +++++++++---
src/main.rs  |  8 +--
 2 files changed, 34 insertions(+)

commit a1b7f3e
Author: human <dev@team>
Date:   Mon Mar 14 08:12:05

refactor cache layer

session: cursor-889
model: claude-sonnet-4
turns: 14
tokens_in: 12,400
tokens_out: 3,200
duration: 4m 22s
attribution: 87% AI

Your team writes code with AI every day. None of that context survives the commit.

The conversations happen. The code lands. Six months later, no one knows who wrote what, why, or how much is AI-generated.

what git sees

Every commit gets a memory.

oobo attaches AI context to every commit: which sessions, which models, which lines, what it cost. Anchors travel with your repo. No external service required.

$git commit -m "switch auth to passkeys"

⚓anchor 9f2c84a created

≈stream captured 11 chat turns

▲delta 7 files touched

⌬dock synced to main

Git stays git. oobo captures the rest.

01

Hooks fire automatically.

Post-commit hooks run in the background. No commands to remember, no plugins to configure. You keep using git exactly as you do today.

02

Sessions are captured.

oobo reads your AI tools (15 supported) and finds which sessions were active at commit time. Read-only. It never touches your tools.

03

Context is stored in git.

Anchors live on a git orphan branch inside your repo. They push when you push. Same permissions, same access control, same infrastructure.

oobo

A living timeline of how your codebase is actually being built.

See which AI tools were involved, tokens burned, human vs. machine. On every commit.

$git commit -m "switch auth to passkeys"

⚓anchor 9f2c84a created

≈stream captured 11 chat turns

▲delta 7 files touched

⌬dock synced to main

$oobo delta HEAD~5..HEAD

AI involvement: 42% → 67% (+25%)

New areas entered: src/payments/, src/webhooks/

Sessions driving: 12 (cursor 8, claude-code 4)

Complexity shift: moderate → high

Summary: Payments rewrite driven primarily by

autonomous agent runs over 3 days.

oobo delta

See how your codebase is evolving.

Compare any two points and see what changed in how the team builds, not just what lines moved.

oobo recall

Your team's AI conversations become searchable institutional knowledge.

Natural language search across every AI session and anchor that produced shipped code. For searching source code itself, use oobo search.

$oobo recall "why did we rewrite the cache?"

⚓anchor 4a1f8c2 (Mar 14)

Session: cursor-session-889 (claude-sonnet-4)

"The redis layer was hitting 200ms p99 under
load. Switched to an in-memory LRU with
write-through to reduce latency..."

2 more results →

$oobo goto t3a8f2c

Restoring workspace to turn t3a8f2c...

Session: claude-code-session-441

Turn 14 of 23 (agent was refactoring auth)

Files at this point: 4 modified, 1 created

$oobo back

Restored to current state.

oobo goto

Jump back to any moment in any AI session.

Travel back to any point in any agent session. See what the files looked like, what went wrong.

oobo blame

See who wrote what. Line by line. Across your entire history.

Every line traced to its origin. Human or AI, which tool, which model, which conversation.

$oobo blame src/auth.rs

●L12 let token = jwt::sign(claims, &key)?;

●L13 set_cookie(res, "session", &token);

○L14 audit_log(&user.id, "login");

●L15 Ok(Response::redirect("/dashboard"))

● AI (cursor, claude-sonnet-4) ○ Human

$oobo mcp install

✓cursor ~/.cursor/mcp.json added "oobo"

✓claude ~/.claude.json added "oobo"

Tools available to agents:

Done. Restart your AI tools to activate.

oobo mcp

Your agents remember what your team already solved.

One command connects Cursor, Claude Code, or any MCP-compatible tool to your engineering memory. Agents get context from past sessions, decisions, and patterns. Automatically.

Benchmarks

Memory compounds. We measured.

Same agent, same tools, same tasks. One gets oobo memory, one doesn't. SWE-bench, Claude Sonnet 4.

15 tools. Zero config.

oobo detects your tools automatically on every commit. No plugins to install, no integrations to configure, no tokens to paste into each tool.

Open source. Local-first. No lock-in.

MIT/Apache 2.0. Single binary, zero dependencies. All data in your repo on a branch you control. No cloud required. Install today, remove tomorrow. Git history unchanged.

View on GitHub

The platform

Your organization's semantic timeline.

The CLI captures data. The platform makes it visible across your entire engineering organization.

Team-wide visibility

See every anchored commit across your organization. Filter by tool, by engineer, by model, by time range. Understand patterns at a glance.

AI usage analytics

Token spend, model usage, AI-authored percentage, autonomous vs. supervised runs. The metrics engineering leaders need, without asking anyone to self-report.

Searchable history

Every AI conversation that led to shipped code, searchable with natural language. Onboard new engineers by showing them how things were built.

Compliance and audit

Know exactly which lines are AI-generated. Export attribution reports. Meet the compliance requirements coming for AI-produced code.

Alignment tracking

Understand whether engineering effort is advancing business goals or drifting. Connect commits to outcomes, not just pull requests.

No workflow changes

Your team keeps using their tools. oobo captures context silently. The platform aggregates it. No forms, no status updates, no friction.

Anchors hold. The codebase drifts.

Start recording which AI wrote what, today. Build engineering memory that compounds with every commit your team makes.

curl -fsSL https://oobo.ai/install.sh | bash

macOS · Linux