--- sidebar: order: 110 title: "Release status" topic: "Reference" resource_type: reference description: "What is available today: public source, invited Cloud beta, public CLI beta, and experimental self-hosting." summary: "The single place that states Tedix availability and release policy" read_when: - Checking what Tedix offers today - Deciding between Tedix Cloud, local mode, and self-hosting - Checking how source, releases, and commits are published visibility: public --- # Release status In short: anyone can read the source and run Tedix on their own computer. The hosted service, Tedix Cloud, is open by invitation. The plugin for Claude, Codex and the ChatGPT desktop app and the `tedix` command-line tool are public betas. Running your own copy on your own cloud account is possible but experimental, with no upgrade or backup guarantee. Beta means it works and we use it every day, but things still change and rough edges remain. The [known limits](#known-limits) below say plainly what Tedix does not check for you. ## Available today | Surface | Status | What it means | | ------------ | ------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Source | Public | [tedix-hq/tedix](https://github.com/tedix-hq/tedix) on GitHub. `main` is the source line; anyone can clone, read and run it. | | Local mode | Preview | From a checkout, `bun run-local` starts OS, onboarding, API/MCP and isolated persistence without model calls. No account is needed. | | Tedix Cloud | Invited beta | The managed service. Access is by invitation; an approved user can create their own organization. Usage beyond the plan allowance is off by default; an organization owner or admin can turn overage on with a monthly cap. | | Tedix plugin | Public beta | Published at [tedix-hq/tedix-plugins](https://github.com/tedix-hq/tedix-plugins) for Claude Code, Codex and the ChatGPT desktop app; Claude chat and Cowork use a custom connector. Devin and OpenCode load the plugin from the source repository. Version 0.2.0 is under review for the Claude directory and the OpenAI plugin directory; ChatGPT on the web is not supported until it is listed. See [Use Tedix in ChatGPT, Codex and Claude](./learning-paths/first-plugin.md). | | Tedix CLI | Public beta | Anyone can download it from [`downloads.tedix.dev`](https://downloads.tedix.dev/latest.json); using it with Cloud needs Cloud access. | | Self-hosting | Experimental | Deployment tooling exists; current `main` is not certified for a fresh account and has no upgrade or backup guarantee. | To request Cloud access, use the [contact page](https://tedix.dev/contact/). Ask questions and share ideas in [GitHub Discussions](https://github.com/tedix-hq/tedix/discussions). Report a security issue privately to [security@tedix.dev](mailto:security@tedix.dev). ## Releases The CLI ships versioned `cli-v*` releases. Product `main` is the source line and fixes land as ordinary commits. There are no nightly, candidate, or long-term-support channels. Tedix Cloud builds from product `main` plus private operational configuration: credentials, live account settings, and runbooks. Public source does not change this deployment boundary. Product code and patches belong in the product repository; operations remain private. ## Commits Maintainers commit directly to `main`. Commit messages and author identities are as public as the files: they name a real person or agent, never a customer, and never carry a credential or infrastructure identifier. The pre-push secret scan checks messages and authors as well as files. Maintainer commits keep their `Work-Item:` and `Agent-Session:` trailers. They identify the maintainer task and the agent session that made the change; the ids do not resolve publicly. Pull requests are currently disabled; see [CONTRIBUTING.md](https://github.com/tedix-hq/tedix/blob/main/CONTRIBUTING.md). ## Known limits - Remembering a conversation is not learning for good. Facts and skills are stored separately, and each run picks the ones it needs. - A completed run records what the tedi reported. It does not check that the reply is correct. - An outside action (a sent message, a created record) is confirmed only by the provider's own reference, not by the run finishing. - Each tedi runs two delegated tasks at a time by default; more wait in a queue. Other delegation limits apply only where an organization policy turns them on. - An agent session's report of success is not verification. A coding lane started by the local dispatcher without a commit on `main` settles as failed, and Work is completed by its owning tedi or person. - Local mode makes no model calls unless you add inference on your own Cloudflare account. It does not configure connectors or deploy anything. - The own-account deployment path has not been re-validated against the current `main`.