v0.52.0is outSee what changed

Open source · Self-hostable · Zero LLM calls in the analysis

Your codebase, explained to your AI and your team.

Repowise reads your repository once and keeps the picture current on every commit. Your coding agent gets the context it was guessing at. Your team sees what a change touches before it merges.

Public repos are free, no account needed.

Or open repowise on its own codebase
For developersFor teams
your agent

› How does auth refresh a session?

get_answer

› Update the docs for auth/

repowise update

› What talks to the billing service?

get_architecture

› Why is the backoff exponential?

get_why

› Simplify the retry in middleware.py

get_risk

› What should we refactor first?

get_health

› Who knows this file?

get_context

› Anything risky in this PR?

get_security
repowise · from the indexLearn moreLearn moreLearn moreLearn moreLearn moreLearn moreLearn moreLearn more
  • Refreshes on 401, then retries with backoff.
  • middleware.py:41–58 · session.py:12
  • Confidence high · 2 sources cited
  • 3 pages rebuilt from the last commit
  • overview · middleware · session
  • Fresh · no model call needed
  • api → billing → stripe adapter
  • 1 import cycle · billing ↔ invoices
  • 4 modules depend on it

  • why it is like this

    The backoff is deliberate. Added after the 2024 outage.

  • Decision 14 · accepted · 2024
  • 7 callers outside this file
  • Usually changes with routes.py
  • 2 tests to run first
  • payments/processor.py · 3.1 / 10
  • Extract the 3 nested branches.
  • 9 bug fixes here in 6 months
  • 2 owners · last fix 3 weeks ago
  • 12 commits in 90 days
  • Bug magnet · read before editing
  • 0 secrets in the diff
  • 1 advisory · requests 2.31
  • Risk low · nothing broken

Measured, not claimed

100k+pip installs

Last 180 days

31.6%fewer agent tokens

django/django at a single pinned commit

0.74bug-prediction AUC

21 repositories, 9 languages, 2,826 files

2.3xmore defects found

2,770 files scored by both tools at the same leakage-free commit

Read the benchmarks

Used by developers at

  • Microsoft
  • Red Hat
  • Siemens
  • AWS
  • Atlassian
  • Accenture
  • NVIDIA
  • SAP
  • Meta
  • Thoughtworks
  • Wix
  • EY
  • Deutsche Telekom
  • Wipro
  • Glovo
  • Google DeepMind
  • Apple
  • OpenAI
  • Intel
  • IBM
  • Oracle
  • Salesforce
  • Adobe
  • Cisco
  • Cloudflare
  • GitLab
  • Spotify
  • PayPal
  • Adyen
  • Revolut
  • Wise
  • Nubank
  • UBS
  • Santander
  • Thomson Reuters
  • Booking.com
  • Trivago
  • ByteDance
  • Tencent
  • LINE
  • Samsung
  • Yandex
  • Arista Networks
  • F5
  • OVHcloud
  • GoDaddy
  • Autodesk
  • Intuit
  • Rippling
  • Gusto
  • SurveyMonkey
  • Scale AI
  • Midjourney
  • Criteo
  • CleverTap
  • TCS
  • Capgemini
  • Endava
  • Nagarro
  • Fujitsu
  • Shopware

Engineers who use Repowise for their projects work at these companies.

Why AI velocity stalls

Fast to write. Slow to trust.

AI made producing a change cheap. Understanding what it touches is still expensive, and someone pays that cost on every task.

01

Agents guess.

Your agent reads three files and confidently changes a function that forty callers depend on. Nothing told it.

02

Humans re-verify.

The diff took seconds to write. The review takes the afternoon, rebuilding enough context to judge it.

03

Teams forget.

The retry logic is shaped by an outage two years ago. The engineer who knew left. Last week an agent simplified it.

One index. Three jobs.

Understand it. Change it safely. Keep it healthy.

Repowise connects structure, history, tests and decisions once. The same evidence then answers three different questions.

docs · auth/middleware.py

Retries

stale updating fresh

Retries the upstream call with exponential backoff.

40def retry(call, tries=3):
41 for n in range(tries):
42 delay = 2 ** n
43 # deliberate: see decision 14
44 raise Upstream("gave up")

why it is like this

The backoff is deliberate. Added after the 2024 outage.

Every commit refreshes the docs

Understand the code

Ask questions with cited answers. Read documentation that rebuilds itself. See why the code is shaped the way it is.

See living documentation
pull request · change risk

repowise bot · risk 6.1

Reaches 6 files. Run the 2 untested callers first.

0 LLM calls per pull request

Change it safely

See every caller a change reaches, the files that usually move with it, and the tests to run, before it merges.

See change risk
code health · payments

defect risk

7.4/ 10

Good · up 0.6

health map

fix next · payments/processor.py

Extract the 3 nested branches. 9 bug fixes here in 6 months.

0.74 bug-prediction AUC across 21 repos

Improve it continuously

Find the files most likely to break next, get the specific refactor named, and remove the code nothing uses.

See code health

Same evidence, every surface

In your agent. On your PRs. In your editor.

Nothing to switch to. Repowise shows up where the work already happens.

How it works

Three steps. One command to start.

  1. 01

    Index once

    Run one command locally, or paste a repository URL. Structure, history, tests and decisions are read into a single index.

    $ pip install repowise && repowise init

  2. 02

    Connect your tools

    Your agent connects over MCP. The GitHub App watches pull requests. The editor extension reads the same index.

    $ repowise serve

  3. 03

    Stay current

    Every commit updates only what changed, in seconds. The picture never goes stale.

    $ repowise update

Who it is for

Two ways in.

For developers

Stop watching your agent read the same file four times.

  • Free, open source, runs on your machine
  • Ten MCP tools your agent already knows how to call
  • Works with Claude Code, Cursor, Codex and any MCP client
Give your agent context

For teams

Put reviewers, leads and agents on one shared picture of the codebase.

  • One index per repo, shared across the team
  • Risk on every pull request, health across the portfolio
  • See how much code AI wrote, and whether it is healthy
See the team plan

Run it your way

Open source. Hosted. Enterprise.

Open source

Index locally, connect your tools over MCP, keep the runtime and the model provider under your control.

View on GitHub

Hosted

Browse public repositories free, sign in for grounded answers, add private repositories when you need them.

Start free

Enterprise

Self-hosted or air-gapped, with the licence, identity and support an organization asks for.

Explore enterprise

What we publish

We also publish the rows we lose.

Every number on this site ships with its sample size, its conditions and the raw data. The benchmarks page includes the measurements where another tool wins.

Read the full record
FREQUENTLY ASKED

Questions, answered

What does Repowise index?

Repowise analyzes source structure, dependencies, call relationships, git history, ownership, code-health signals, tests, and engineering decisions. It turns that evidence into a reusable codebase model and generated documentation.

Can I browse without signing in?

Yes. You can open and browse supported public repositories without an account or installation.

What do I get when I sign in?

The Free hosted experience includes 10 grounded answers each month. Those answers use indexed repository evidence rather than treating the codebase as an unstructured prompt.

Does Repowise replace my coding agent or IDE?

No. Repowise supplies codebase context to developers and existing tools through the hosted application, CLI, and Model Context Protocol. Your agent or editor remains where the work happens.

Can I run it myself or inside my organization?

Yes. The core engine is open source under AGPL-3.0 and can run locally with your model provider or a local model. Repowise is also available as a hosted product, with an enterprise path for organizational deployments.

Start with a real repository

Index a repo free.

Paste any public repository, or install the PR bot on your own.

Explore a repositoryInstall the PR bot