Cascade

Meeting recording in. Tested pull request out. Self-hosted. Your LLM key. Your code never leaves your org.

Three on-ramps, one pipeline

Start from a meeting

Drop in an mp3 or mp4. Cascade transcribes, identifies speakers, extracts user stories, and walks you through review before any code gets written.

Start from a ticket

Point Cascade at a Jira, Linear, GitHub, GitLab, or Azure Boards ticket. It pulls the requirements and builds a PR with tests.

Start from a prompt

Type one line. Get a plan, code, tests, and a PR. Ideal for the small features that pile up on every team’s backlog.

Why Cascade exists

Most AI dev tools assume your team works exactly the way they expect. You code in their IDE. You pay for their LLM. You host on the VCS they support. You start work by typing into a chat window.

Real engineering teams don’t look like that. Some teams discuss requirements in standups. Some have a corporate Copilot subscription and no Anthropic budget. Some can’t legally send code to a SaaS at all. Some are on GitLab. Some are on Azure DevOps. Some still use Bitbucket.

Cascade was built to meet teams where they already work.

What makes it different

Self-hosted by default

Cascade runs on your machine or your CI. Code never leaves your org except to the LLM provider you chose.

Bring your own AI

Anthropic, OpenAI, Google Gemini, Ollama, or your existing Claude Code subscription. No new API key required if you already have one.

Polyglot from day one

Python, TypeScript, JavaScript, Go, Rust, Java, Ruby, C#. Adding a ninth language is a single registry entry.

Humans at every gate

Cascade plans, codes, and tests. Humans review and approve. Built for trust, not autonomy theatre.

Team memory baked in

A team-memory directory captures your conventions, decisions, and prior work. Every AI stage reads from it.

Free forever (MIT)

Use it commercially, fork it, ship it inside your enterprise. No license keys, no usage caps.

Get started in 60 seconds

pip install cascade-agent
cascade init
cascade configure llm anthropic --key sk-ant-xxx --set-default
cascade configure vcs github --token ghp-xxx
cascade prompt "Add cursor pagination to /api/users"

That’s the full setup. The last command runs the whole pipeline: plan, generate code, generate tests, run them, open a PR.

Read the full getting-started guide →

Status

Pre-alpha, building in public. Soft launch target: 2026-09-15.

If you want to be part of the early beta, drop a line.