Tedix: Hand Off Work to AI Assistants and Get Back Only the Decisions That Need You

Who it's for

Tedix is for people who use AI assistants regularly, want to hand routine work over wholesale, and step in personally only at the key decision points. The official positioning: once the work is handed off, Tedix brings back only the decisions that need you, ranked by what matters most, and handles the rest. In other words, the question it answers is not whether AI can be invoked, but which items deserve your own eyes. If you already push tasks forward inside Claude, ChatGPT, or Codex and want to reduce the back-and-forth of checking progress, it is worth a try.

On clients, official support covers Claude, ChatGPT, and Codex; other clients that speak the MCP protocol connect through the MCP gateway. Whether assistants outside this range can connect smoothly is not stated in the evidence and needs your own verification.

The prerequisite for getting started is a local toolchain: Git, Bun, and Node.js 22+, with an official emphasis on a real Node, not a Bun shim. This local path needs no account and no cloud credentials; the cost of trial and error lies mainly in preparing the environment. Whether it fits the kind of work in your field is likewise not covered by the evidence, so verify with low-risk tasks first.

Not a fit for now: users who only want a ready-made cloud service and refuse to install a local toolchain. Tedix Cloud is not open for sign-up yet and works by invitation; without an invitation and without a local environment, there is no way to try it at this stage. Teams with hard requirements for cloud capabilities are advised to secure an invitation first and evaluate afterwards.

A first try in 3 minutes

The recommended primary path is the local demo mode: it runs the app locally with sample data, does not depend on the cloud, and needs no account or cloud credentials.

Step one, confirm the prerequisites: install Git, Bun, and Node.js 22+. If this set of dependencies cannot be installed locally, this path will not work.

Step two, clone the repository and start the demo:

git clone https://github.com/tedix-hq/tedix.git && cd tedix
bun run-local --demo

Step three, watch for the success signals: the app starts locally and loads sample data. The official description stops there; which operations the demo interface actually offers and what the sample data contains are not stated in the evidence, so observe and confirm locally. For the first try, it is advisable to only observe and not submit real tasks, going deeper once the behavior matches expectations.

Two reminders: first, this step is a local run and involves no cloud credentials; second, whether the app reads other local data or connects to the network is not stated in the evidence, so do not assume it has no impact. Neither point blocks the first attempt, but both affect whether you keep using it.

The other official access methods are secondary references only, not peers of the primary path: agents that support Agent Skills can run npx skills add tedix-hq/tedix-plugins to add Tedix's skills; beyond the officially named clients, clients that speak MCP connect through the MCP gateway. What configuration each of these methods needs is not given in the evidence.

Use cases

Case one: stay off the cloud and use sample data to learn how its decision mechanism works. Scenario: you want to judge whether the working style of “bringing back only the decisions that need you” suits you, without providing any real data. Approach: complete the clone above and run bun run-local --demo. Expected result: the app starts locally with sample data, with no account or cloud credentials at any point; which operations the demo offers and what output you get are not listed in the evidence, so treat what you actually see locally as the standard.

Case two: add Tedix skills to an agent you already use. Scenario: the agent you use daily can read Agent Skills, and you want it to gain Tedix's work-handling capability. Approach: run npx skills add tedix-hq/tedix-plugins. Expected result: officially, any agent that reads Agent Skills can add Tedix's skills this way; which skills are actually provided after adding, and how each is invoked, are not given in the evidence and cannot be confirmed down to executable detail, so check after adding and do a trial run with an inconsequential task first.

Boundaries of use

• Read vs. write: among the specific capabilities exposed by demo mode and the MCP gateway, which are read-only and which write local or remote state is not stated in the evidence; clarify the impact before executing an action, and a local write should not be treated as read-only either.
• Credentials and data: for the local primary path, the official statement is explicit that no account and no cloud credentials are needed; the scope of local files the app reads, whether it connects to the network, and how data is handled are all unconfirmed.
• Cloud gap: Tedix Cloud is invitation-only and not open for sign-up; which capabilities the local demo and the cloud service each cover, and how wide the gap is, is not stated in the evidence.
• Client consistency: whether the officially named clients and the clients connected through the MCP gateway offer the same features is unconfirmed.
• Undo and retention: the evidence does not mention any undo path for operations, usage quotas, or data retention; the absence of a statement does not mean the absence of limits, so defer to the official documentation before critical operations. The scope of the sample data content is likewise unstated.