Codegen

Codegen runs agents inside isolated environments where they can execute code, install dependencies and test their own changes without touching a developer's machine or a shared build box. The sandbox is the safety boundary that makes autonomous work practical: the agent can try things, see failures and iterate before anyone reviews the result. This matters for enterprise adoption, where the alternative is giving an agent broad access to production systems.

GitHub is how Codegen reaches a repository and performs every git operation. Agents create and manage pull requests, help resolve merge conflicts, run code reviews, search a codebase and sync branch changes. The GitHub App, published as codegen-sh, is the one-click install that grants these permissions, and the integration covers everything from opening a PR to reporting CI results.

Codegen meets teams where they already work. Agents are triggered by assigning a ticket or mentioning them in Slack, Linear, Jira, ClickUp or Monday.com, and each ticket or thread gets its own dedicated agent that follows the full conversation. That design means work stays tied to the original request, and updates flow back into the tracker or channel instead of a separate tool.

Code review is a first-class feature rather than an add-on. The PR review agent reads a proposed change, checks it against the project and leaves comments before a human reviewer spends time on it. Teams use it to catch obvious problems early and to keep review latency down as agent-generated pull requests start to outnumber human ones.

The platform lets an organization pick which large language model powers its agents. You can choose among major providers, configure custom API keys and base URLs, or connect self-hosted models if you have a specific arrangement. Settings apply globally across the organization, with per-repository or per-agent overrides where a plan supports them, so governance stays centralized.

For buyers with strict data requirements, Codegen ships as a cloud-native Kubernetes application that runs inside your own infrastructure. It supports Docker images and AWS AMIs, and it can use your own AI models and API keys end to end. This is aimed at teams with data sovereignty rules, air-gapped environments or compliance mandates that keep all code inside corporate boundaries, and it is offered on the Enterprise tier.

Codegen exposes a REST API under api.codegen.com and a Python SDK for programmatic control. Teams use it to kick off agent runs from automated workflows, wire agents into internal dashboards and build custom tooling around the platform. The API takes organization-scoped requests with a bearer token, and it is the path for anything the built-in integrations do not cover.