Code
GitHub and GitLab repositories, files, branches, pull requests, and issues. The AI reads what already exists before anyone writes a requirement.
Specs that know your codebase
Your codebase is already a spec. A PRD that ignores it can sound complete and still be wrong about your own product. Kairo connects product work to your repositories, designs, and team channels, then serves finished specs to your coding agents through its MCP server.
Context in
The context your specs are missing
Design
Import Figma designs into Kairo documents, so product, design, and AI reason from the same screens.
Documentation
Upload the documents your team has already written and Kairo uses them as context for your specs. Connectors for Confluence, SharePoint, and Notion are coming soon.
Chat
For what was already said elsewhere, Kairo brings context in from your Slack conversations.
GitHub 
GitLab 
Figma 
Slack 
Confluence Soon 
SharePoint Soon 
Notion Soon 
Teams Soon 
For the whole team
Ask the codebase, in plain language
With a repository connected, product owners, product managers, and domain experts get engineering-level answers in seconds: what exists, what a change would touch, which constraints apply. No engineer has to be pulled off their work to answer.
- "What happens today when a payment fails?"
- "Where are invite permissions checked?"
- "Does anything already do this?"
Context out
Specs served to any agent over MCP
The MCP server is the main way to get specs out of Kairo. Coding agents read the change request directly: its definition, its plan, and its linked documents. Nobody copies and pastes prompts.
- Agents get the same context your team agreed on
- Your specs stay in Kairo, not in one vendor's agent
Hands off via MCP to Claude Code Copilot Codex Cursor
Your agents will change. Your decisions shouldn't be trapped in them.
Security and control
Connected, and still in control
Connecting your code and your tools raises one question: who can see what? Kairo connects through secure OAuth and service accounts, so access belongs to the team, not to one person's credentials, and follows clear rules for people and for AI.
Workspaces, projects, and roles
Each project has its own documents, chats, change requests, skills, and AI team. Owners and admins manage members, roles, and integrations.
Membership-gated access
People only reach the projects, documents, and chats they belong to. The same rule applies to AI assistants.
AI that stays in bounds
AI assistants act only with the context, tools, and permissions granted for each turn. Their edits arrive as suggestions your team accepts or rejects.
Your data stays yours
Built and hosted in Europe, with GDPR-compliant data handling. Your data is never used to train any AI model, ours or anyone else's.
Give your specs the context they're missing
Join the pilot and see what your specs look like when they know your codebase.