The setup process appears straightforward, and the concept of a single project memory for all AI tools is well executed.
Haven't used AI-Continuum yet?
Ask Trustburn
Questions about AI-Continuum
What is AI-Continuum?
AI-Continuum is an alpha software tool that provides a local-first shared project memory for AI assistants, operating as an MCP-native server compatible with clients like Claude, Codex, and Cursor. It allows users to read and write to a common project memory including goals, active work, blockers, and decisions, emphasizing user control through visible write receipts and one-command undo functionality.
How does the memory storage and engine operation work?
The platform stores memory in plain files, SQLite, and Git, and is distributed as alpha software under the name ai-continuum (formerly ai-os). It is part of the aihq suite and operates as a hosted service where the engine runs on their host to index and read memory with outside AI models, starting with Claude and falling back to OpenAI's Codex if needed.
What are the steps to set up and connect machines?
Users can sign in with Google to forge their vault into a repository they own, then run the setup from the wire on each machine to approve it in their browser. Each machine gets its own key, and the coding agent's plugin keeps each session as it ends, allowing users to stop every machine at once from their keys page.
How does the claim admission and briefing process function?
Claims arrive as proposals and wait for users in 'needs you' unless the portal itself confirmed the document a claim quotes. Users can admit, correct, or strike claims in the edition, and from then on, Claude Code starts each session briefed on what is admitted, while Codex gets no briefing yet.
What are the current limitations and alpha features?
The system operates as alpha software that is hosted and free to use, with memory starting empty where claims appear only as they are proposed and admitted. Every act on a claim can be undone, and nothing reaches an agent until it is admitted by the user or the portal after checking the document the claim quotes.
How does data export and vault deletion work?
Users can export their records, including claims, records, documents their machines sent, and corrections they made. Deleting the vault closes it at once with a thirty-day undo period, after which the team purges it by hand and stops every key, though the engine retains data out of reach until a specific removal is built.
What is the current Trustburn rating and review count?
The Trustburn rating is 4 out of 5 from 5 reviews, with the platform emphasizing user control through visible write receipts, authority labels, history, and one-command undo functionality. The project is distributed as alpha software and is part of the aihq suite.
What do reviewers praise about the platform?
Some reviewers praise the emphasis on user control through visible write receipts, authority labels, history, and one-command undo functionality. The platform allows users to forge their vault into a repository they own and run the setup from the wire on each machine to approve it in their browser.
How does the system manage agent sessions and claim waiting?
The platform allows users to stop every machine at once from their keys page, and every briefing an agent is given leaves a receipt. Claims that an agent saves wait in 'needs you' until the user admits them, ensuring nothing reaches an agent until it is admitted.
What is the official website of AI-Continuum?
The official website of AI-Continuum is ai-continuum.dev.
Post your question publicly — other visitors, and whoever manages this page, can answer. It will be visible on this page.
Want to hear about this company again? Follow it and we’ll email you when a new review appears.
Based on reviews, listing data and the company's website. Verify anything important with the company directly.
5 Reviews
While the product is free to try, the website does not clearly outline long-term pricing for hosted convenience or enterprise features. This lack of transparency regarding future costs is a concern for teams looking to adopt the tool permanently.
From the information available, the local-first design seems robust for developers who prioritize data ownership. The fact that the memory lives in plain files and git is a strong selling point. Since it is still in alpha, one should expect some moving parts, but the undo and veto features provide a good safety net. It looks like a viable solution for keeping project context consistent across different AI clients.
I have been frustrated for years by having to re-explain my project context to every new AI tool I try. AI-Continuum solves this by giving all my AIs one shared memory in files I actually own. It feels like a breath of fresh air to see Claude, Codex, and Cursor all reading the same living memory without me losing track. The local-first approach is a huge win because my data stays on my machine with zero default telemetry. I really appreciate the transparency about this being alpha software; you know what you are getting. The ability to veto changes and see every write with an immutable receipt gives me the control I missed in other tools. It is not perfect yet, but the direction is exactly what developers need.
The unified project memory works well across clients, though being alpha means occasional sharp edges during daily workflows.
Questions & Answers
No questions yet — be the first to ask.
Sign in to post
Your text is saved — you'll come right back to post.
or with email
Don't have an account? Create one
News & Press
No news yet. This is where the business shares its own updates and press.
Photos
No photos yet. Photos shared by reviewers will show up here.
About AI-Continuum
AI-Continuum provides a local-first software tool that gives AI assistants a shared project memory stored in files on the user's machine. It operates as an MCP-native server that works with MCP-compatible clients such as Claude, Codex, and Cursor, allowing them to read and write to a common project memory including goals, active work, blockers, decisions, and open questions. The platform emphasizes user control through visible write receipts, authority labels, history, and one-command undo functionality, with memory stored in plain files, SQLite, and Git. The project is distributed as alpha software under the name ai-continuum (formerly ai-os) and is part of the aihq suite, with installation available via an npm command and source code hosted on GitHub.
- Website
- ai-continuum.dev
