Early access.
There is no public download yet. Kontinuity is in early access, and we let developers in by request.
What you need.
- A computer running Linux, macOS or Windows.
- A coding agent that can run a command, or a desktop that speaks MCP.
- An account for Cortex, which holds your memory and which Keon runs.Sign-in not open yet
What works now, and what does not.
We say both in the same place, so you do not have to find out by trying.
Works in our own use
- A briefing of what you confirmed, given to an agent at the start of a session.
- Said, guessed and confirmed kept apart, with a review queue for guesses.
- A task board that agents claim work from.
Not yet
- A public installer.Not published yet
- Sign-in for new accounts on the hosted memory service.In development
- Desktop builds verified on all three systems.In development
- Signed receipts for each memory write.In development
- Hosted agents that cannot run a command, such as hosted Grok Bots.Not supported
Which harnesses connect.
If it can run a command or speak MCP, it can join. This is how each one connects today.
| Harness | How it connects | Briefing at the start of a session |
|---|---|---|
| Claude Code | Agent Skill and command line | Automatic with the optional startup hook; otherwise the skill loads it. |
| Codex | Agent Skill and command line | The skill tells the agent to load it. No startup hook. |
| Gemini CLI | Agent Skill and command line | The skill tells the agent to load it. No startup hook. |
| Grok Build | Agent Skill and command line | Automatic with the optional startup hook; otherwise you load it. |
| Cursor | Agent Skill and command line, or the MCP server | The skill tells the agent to load it. No startup hook. |
| pi | Extension | Built in. |
| MCP desktops | MCP server | The server hands the agent the briefing. |
Platform notes.
Linux is where we test most.
macOS runs in our automated tests.
Windows is not fully verified yet. If you use Windows, expect rough edges and tell us what you find.