Code Autopilot
An AI coding agent inside GitHub, not a separate app. Describe a task in an issue and it replies with a plan and code, opens a pull request, or reviews an existing one. For solo developers and small teams.
What is Code Autopilot?
Code Autopilot is an AI development agent grafted onto GitHub and installed as a GitHub App, rather than a product with an interface of its own. The site sums it up as ChatGPT inside your Github account, and its page title calls it a Github GPT companion.
The trigger is an issue. The tool listens to the issues created on a repository and answers in the thread with a proposed solution: an implementation plan, code snippets to copy and paste, or a pull request opened in one click. The publisher describes it as working on top of an existing codebase, solving issues, replying to comments and reviewing pull requests. Four uses are pushed forward: chatting inside issue and pull request threads, implementing a feature, fixing a bug, and speeding up review by summarising what changed and explaining what the author meant.
Its scope goes beyond one repository. After installation it indexes the code, then navigates the whole codebase and several repositories belonging to the same organisation. The engine is described only in general terms — coding agents driven by state-of-the-art LLM models — with GPT-4 named on the Starter plan and snippets sent to OpenAI’s API. The Enterprise plan replaces that arrangement with self-hosted private LLMs and unlimited issues. Compatibility is claimed for virtually any programming language, and the homepage claims more than 1,000 GitHub users.
Getting started is presented in three steps: install the GitHub App, create an issue, let the tool index the repository and reply. Billing is per organisation, and every repository under that organisation draws on the same quota. The product also has an open source ancestor: the autopilot project by fjrdomingues, whose README points to codeautopilot.com as the hosted GitHub App version.
Two observed limits deserve stating plainly. The maintenance signals conflict: the GitHub App is live and installable, but the install button on the site returns a 404 and the documentation subdomain no longer resolves in DNS. And nothing contractual is published at all — no terms of service, no privacy policy, no data processing agreement, and no mention of the GDPR.
What it does
- Propose a code solution from the description of a GitHub issue
- Suggest a fix for a bug reported in a ticket
- Open a pull request in one click from the proposed solution
- Review a pull request: comment on it, explain the author’s intent, suggest improvements
- Summarise the changes contained in a pull request
- Explain how an unfamiliar codebase works
- Discuss and refine a solution in issue or pull request comments
When to use Code Autopilot / When not to
A quick filter to help you decide if Code Autopilot is the right fit.
When to use Code Autopilot
- Solo developers and small businesses whose work is already tracked as GitHub issues — the Individual plan is positioned as “perfect for individuals or small businesses”.
- Teams that want to keep their existing development workflow intact, including organisations spread across several repositories the agent can navigate.
- Pull request reviewers buried under changes, who need a plain summary of what a branch actually does before reading the diff.
- Less technical contributors who describe an outcome rather than write it: the site claims no advanced dev skills are needed.
- Engineering organisations that cannot let code leave their infrastructure and can afford the Enterprise plan, which offers self-hosted private LLMs.
When not to use Code Autopilot
- Anyone whose code is not hosted on GitHub: no other forge is supported, and GitLab and Bitbucket are never mentioned.
- Teams that refuse to send source code to a third party — on the Starter and Individual plans, snippets are forwarded to OpenAI’s API.
- Buyers who need a written commitment: no terms of service, no privacy policy and no data processing agreement are published.
- Developers looking for an API to embed elsewhere, or a web, mobile or desktop application: everything happens inside GitHub comments.
- Anyone needing round-the-clock support or heavy volume for free: support is “with the founder” at a single address, and the free tier caps at roughly five issues.
How to use Code Autopilot
A typical end-to-end flow, from setup to results.
- Install the Code Autopilot GitHub App on the repositories you choose, from github.com/apps/codeautopilot
- Note that the install button on the homepage returns a 404, so go through the App page directly
- Let the tool index the repository
- Create an issue describing the task to implement or the bug to fix
- Read Autopilot’s reply in the thread: an implementation plan and the matching code
- Refine the answer by replying in the comments of the issue or of the pull request
- Copy and paste the snippets into your branch, or open a pull request in one click
- Let Autopilot comment on and summarise existing pull requests to speed up review
- Point it at other repositories of the same organisation once the workflow suits you
- Pick a plan on GitHub Marketplace when the free allowance runs out; billing is per organisation
Pros & Cons
Pros
- No tool switching: everything happens in the GitHub issues and pull requests the team already uses
- Three-step setup, with no API key, no configuration and no file to add to the repository
- A permanent free plan of roughly five issues, with no card announced
- Low paid entry point at 19 USD, billed per organisation rather than per seat
- Covers the whole chain, from issue to code to pull request to review
- Self-hosted private LLMs on the Enterprise plan, for teams whose code cannot leave their infrastructure
- Works with virtually any programming language, and the original open source component (fjrdomingues/autopilot) can be inspected
Cons
- No privacy policy, no terms of service and no data processing agreement are published
- No mention of the GDPR anywhere on the site
- The only install button on the homepage points to a GitHub Marketplace listing that returns a 404; the working App lives at github.com/apps/codeautopilot
- The documentation subdomain docs.codeautopilot.com has no DNS record, although the site and the App page still link to it
- Locked to GitHub: no other code forge is supported
- On every plan below Enterprise, code snippets are sent to OpenAI’s API
- Thin support and quiet maintenance signals: support is “with the founder” at one email address with no response commitment, the docs repository was last pushed in October 2024, and the footer copyright stops at 2025
Pricing & Plans
A permanent free plan is available: the Starter tier costs nothing and covers roughly five issues. The cheapest paid tier is Individual at 19 USD, which raises the allowance to about thirty issues per month. The highest tier, Enterprise, is listed at 300 USD and adds unlimited issues and self-hosted private LLMs. Billing is per organisation: all repositories belonging to the same GitHub organisation draw on a single quota. No paid trial period is announced, the free plan serving that role. The amounts are displayed without a time unit beside them; the monthly cadence follows from the “issues/month” allowance and from the plans being sold through GitHub Marketplace.
- Starter — Free — around 5 issues
- implement
- fix and ask questions
- pull request analysis
- access to GPT-4
- Individual — 19 USD — around 30 issues/month
- same capabilities
- support with the founder
- Enterprise — 300 USD — self-hosted private LLMs
- unlimited issues
- same capabilities
- support with the founder
Data, GDPR & hosting
A consolidated view of how Code Autopilot handles your data.
GDPR overview
There is no GDPR mention whatsoever. Across the thirteen files collected from the site and its public pages, the terms GDPR, CCPA, SOC 2, ISO 27001, HIPAA and compliance do not appear once. No privacy policy, no trust or security page and no data processing agreement is published. No EU representative is designated under Article 27, no legal basis is stated, no retention period is quantified, and no procedure is described for exercising access, rectification or erasure rights. The silence is notable, since the publisher’s verified GitHub organisation places it in Portugal and therefore within the scope of the regulation. The only processing operation disclosed anywhere is the FAQ statement that code snippets are sent to OpenAI’s API.
Who owns the data?
Nothing contractual settles ownership, because the site publishes neither terms of service nor a privacy policy. The only statements on the subject are four homepage FAQ answers. They say the publisher does not store your code, keeps a temporary copy only while an issue is being resolved, and afterwards retains metadata about the repository rather than the source code. They also say snippets are sent to OpenAI’s API and that OpenAI does not use them to train its models. No intellectual property clause covers your inputs or the generated code. The GitHub App listing refers to separate third-party terms and a privacy policy, without linking to them — and neither document exists at a public address.
Reuse rights
No document states what you may do with the output, since there are no terms of service and no privacy policy to consult. In practice the result lands in your own space: code arrives as snippets posted in a comment for you to copy, or as a pull request opened on your own repository through the GitHub App you installed, which stays yours throughout. No licence, no restriction and no assignment of rights is published in either direction. That silence is worth reporting rather than interpreting: the publisher has neither granted you explicit rights over the generated code nor claimed any over it, so anyone with a real reuse constraint should get the point confirmed in writing before relying on it.
Data retention & training
Hosting summary
No hosting jurisdiction is published: no country, no region and no provider is named anywhere. The domain resolves to 162.159.140.98, an anycast address belonging to Cloudflare (AS13335). That tells you the site is served through a content delivery network and nothing about where data is stored: an anycast address answers from whichever edge node sits closest to the visitor, so geolocating it identifies a delivery point, not a storage location. The only data flow the publisher documents is the transfer of code snippets to OpenAI’s API, mentioned in a single FAQ answer. The publisher declares Portugal on the GitHub organisation that GitHub has verified for this domain, but that is a profile field, not a hosting commitment. No subprocessor list, no data processing agreement and no international transfer clause is published. The one configuration that avoids a third party altogether is the Enterprise plan, whose self-hosted private LLMs move processing onto the customer’s own infrastructure.
Things to keep in mind
Risks and trade-offs to weigh before adopting Code Autopilot.
- Source code leaves your perimeter: snippets are sent to OpenAI’s API on every plan below Enterprise, with no published processing agreement to frame it
- With no terms of service and no privacy policy, there is no written recourse if an incident occurs
- No GDPR mention at all, although the publisher declares itself in Portugal and therefore inside the European Union
- Double dependency: on a single forge, GitHub, and on a very small publisher reachable at one email address
- Uncertain longevity: documentation offline, install link broken, little public activity since late 2024
- AI-generated code merged without genuine review accumulates technical debt and hides subtle bugs
- Handing debugging over as a reflex erodes the team’s own grasp of its codebase, and a per-issue quota makes cost unpredictable when threads go back and forth
Setup & Integrations
Technical difficulty
Low in principle. Three steps are announced: install the GitHub App, create an issue, read the answer. No API key, no configuration and no file to add to the repository are described, and the site claims no advanced dev skills are needed. The real prerequisites are a GitHub repository and the right to install an App on the organisation. Two frictions were observed: the install button on the homepage returns a 404, so installation has to go through github.com/apps/codeautopilot, and the help documentation is unreachable because docs.codeautopilot.com has no DNS record.
Deployment
Integrations
Behind Code Autopilot
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Where does Code Autopilot actually run?
Which programming languages does it handle?
Where does my code go?
Is my code stored?
How many repositories can I install it on?
Is there a free plan?
Can I avoid sending my code to a third party?
Are there terms of service or a privacy policy?
Is there an API, or a mobile or desktop application?
How do I contact the publisher?
Should you pick Code Autopilot?
Code Autopilot makes a narrow, legible promise: put a coding agent where the work is already discussed, in the GitHub issue. Describe a task or a bug and the tool answers in the thread with a plan and code, then opens a pull request or comments on an existing one. Nothing to install locally, no configuration, no second interface to learn. The commercial terms are just as plain: a permanent free tier of roughly five issues, a first paid tier at 19 USD, billing per organisation rather than per seat, and self-hosted private LLMs reserved for the 300 USD Enterprise plan.
The counterpart is substantial. Not one contractual document is published — no terms of service, no privacy policy, no data processing agreement, and no mention of the GDPR — while code snippets are sent to OpenAI’s API on every plan below Enterprise. The lifecycle signals conflict too. The GitHub App remains live and installable, but the only install button on the site leads to a Marketplace listing that returns a 404, the documentation subdomain no longer resolves, and public activity has been quiet since late 2024. Nothing announces a shutdown; nothing confirms active development either.
The reading that follows is straightforward. For a solo developer or a small team already living in GitHub issues, on code whose confidentiality is not critical, the tool is cheap to try and costs nothing to drop. For professional use on proprietary or regulated code, the missing paperwork is not a detail: there is no written recourse if something goes wrong, and no documented way to keep the code away from a third party below the Enterprise tier. Check those points directly with the publisher before committing anything that matters.
- Choosing a selection results in a full page refresh.
- Opens in a new window.