Tejus documentation

Tejus is a local-first desktop workspace for coding agents on Windows. It brings agent conversations, tasks, terminals, browser work, an IDE, diffs, Git worktrees, handoffs, automations, and pull-request delivery into one application.

What is Tejus?

Tejus gives you two ways to run coding agents:

• The built-in Tejus agent. Nothing to install. Sign in with Google to use Tejus Cloud models with a free daily token allowance, or add your own API keys (OpenAI, Anthropic, Google Gemini, DeepSeek, and others) and use those models without signing in.
• Coding-agent CLIs you already have, such as Claude Code, Codex, Cursor, OpenCode, Antigravity, Grok Build, Devin CLI, Factory Droid, Pi, and Oh My Pi. For these, the provider still owns its account, models, tools, permissions, and service.

Either way, Tejus owns the durable task, working environment, transcript, review surfaces, and delivery workflow around the agent session.

What's in Tejus 1.0

• The IDE: open any chat in the Editor view with a file explorer, editor tabs, search, source control, terminals, and the chat beside your code.
• Tejus Cloud: Google sign-in, free daily cloud tokens, your own API keys, and Plan & Usage, all in the Tejus account group of Settings.
• Security review: after an agent turn, Tejus scans the files it changed for secrets, injection, and other risky code.
• Background tasks: a panel per chat that lists what its agent has running, such as shell commands, subagents, and workflows, with a stop button, plus local servers you can open in the in-app browser.

What does Tejus control?

• Projects and tasks: keep several repositories and durable conversations available without mixing their context.
• Agents: select the Tejus agent or a supported coding-agent CLI, its model, and provider-specific options from one interface.
• Development surfaces: use terminals, browser tabs, the IDE, files, diffs, and Git tools beside the conversation.
• Git environments: work in the local checkout or isolate concurrent tasks in linked worktrees.
• Review and delivery: inspect changes, verify behavior, commit, push, and create pull requests.
• Orchestration: use provider handoffs, automations, Agent Gateway, and restricted External MCP access where appropriate.

Does Tejus include model access?

Yes, optionally. The built-in Tejus agent can use Tejus Cloud models after you sign in with Google and activate your account by e-mail; the free plan includes a daily allowance of cloud tokens. You can also add your own API keys under Settings → Your API keys.

You do not need Tejus Cloud to use Tejus. Coding-agent CLIs keep using the accounts, subscriptions, and API credentials already configured on your machine, and their model availability and usage limits remain provider-owned.

Start with Installation, Tejus Cloud, and the Provider guides.

Does Tejus upload code to its own cloud?

Your projects, chats, settings, and workspace database stay on your machine, under %USERPROFILE%\.tejus. Tejus's backend only receives the following:

• Your Tejus account, if you sign in: the Google account you sign in with, your plan, and your daily Tejus Cloud usage.
• Tejus Cloud model requests: when the Tejus agent uses a Tejus Cloud model, its requests (your prompt and the conversation, file content, and tool results the agent sends to the model) pass through Tejus's hosted proxy to the model provider. The proxy also records token usage against your daily allowance.
• Usage counts for your own API keys: requests made with your own keys go directly from your machine to that provider. When you are signed in, Tejus uploads one usage row per turn (provider, model, input, output, and cached token counts, the chat's ID, and a timestamp) so Plan & Usage can show it. No prompts or code are uploaded. Your installation ID, OS, and app version are attached only if analytics are on.
• Analytics, only if you opt in: under Settings → General → Privacy → Share anonymous usage analytics. This covers app version, OS, launches, and which features and models are used, never prompts, code, file paths, or keys. Events carry a random installation ID, and your account sign-in when you are signed in.
• Feedback and crash reports, only when you send them: see Report a problem.
• Update checks: the app checks Tejus's update feed for new versions.

Coding-agent CLIs such as Claude Code or Codex talk to their own services. Tejus is the control surface for them, not a proxy. The provider you select still receives the prompts, file snippets, diffs, terminal output, or tool results it needs for that session.

See Privacy for the public data boundary.

How should I run my first task?

Use one clean Git repository, one provider you have already verified, and one bounded objective. Follow the transcript, inspect the complete diff, run the relevant checks yourself, and commit only the changes you understand.

Use the Quickstart for the shortest path or Your first task for the complete walkthrough.

How do I run several agents safely?

Give each body of work one owner and use separate worktrees for concurrent writers. Keep objectives independently reviewable, preserve Git state before integration, and assign one integration owner.

Continue with Parallel agents and Worktrees.

Where do I go when something fails?

Start with the symptom-specific Troubleshooting hub. Preserve work before restarting or cleaning up, identify the failing layer, change one variable at a time, and collect the smallest useful evidence.