teta.so
Agent

An AI developer inside your editor.

Teta Agent reads, writes, and runs your code. It generates images, takes screenshots, debugs issues, and follows structured workflows — all through natural conversation.

Full Code Control

Read, write, edit, and search across your entire project. The agent creates files, modifies components, refactors code, and runs regex searches — with the same tools a developer uses.

Read Write Edit Glob Grep

Terminal Access

Run any shell command. Install packages, run build scripts, execute tests, check logs. The agent operates the terminal just like you would — with full output visibility.

agent-terminalagent@teta:~$ npm install d3added 120 packages in 2.5sfound 0 vulnerabilitiesagent@teta:~$ git commit -m "Add d3 charts"[main 8f3b1a2] Add d3 chartsagent@teta:~$ _

Image Generation

Powered by Gemini

Generate images from text prompts directly inside your project. Icons, hero images, illustrations, backgrounds — saved to your codebase and ready to use.

"A beautiful mountain landscape..."

Visual Debugging

The agent can take screenshots of your running app and read browser console logs. It sees what you see — errors, layout issues, runtime bugs — and fixes them in context.

Screenshots Console Logs
► Overlapping layout detected in .hero-section► Agent applying fix: justify-content: space-between► Layout verified via visual regression.

Smart Workflows

Built-in skills for structured development. The agent brainstorms before building, writes plans, follows TDD, debugs systematically, and verifies before marking work done.

/brainstorm /plan /debug /tdd /verify /design
/plan feature_x1. Research2. Write Tests3. Implement4. Verify

Web Access

Search the web and fetch documentation in real time. The agent looks up APIs, reads docs, and finds solutions — no copy-pasting needed.

https://docs.api.com/v2

Interactive Collaboration

The agent asks clarifying questions when it needs your input. Multiple choice, step-by-step wizards, open-ended — it pauses, asks, and continues based on your answers.

Should I use Tailwind or plain CSS?Let's use Tailwind v3.Got it. Configuring Tailwind...

Supervised Mode

Review and approve every code change before it happens. In supervised mode, file writes, edits, and terminal commands require your explicit approval. Full control when you need it.

agent@teta:~$ rm -rf ./old_build⚠️ Action requires user approval...Action approved. Deleting directory.agent@teta:~$ _Permission Requiredrm -rf ./old_buildApprove Action

MCP Servers

The agent connects to external tools through the Model Context Protocol. Image generation, browser automation, documentation lookup, database access — all plugged in as MCP servers that the agent discovers and uses automatically.

teta-image-gen teta-browser-tools context7 supabase

Sub-Agents

The agent spawns lightweight sub-agents to handle parallel tasks. Research, file exploration, code review — each sub-agent works independently and reports back, keeping the main conversation focused.

Superpowers Skills

A library of structured development skills the agent can invoke. Each skill defines a rigorous workflow — from brainstorming to debugging to code review — ensuring the agent follows senior-level engineering practices.

brainstorming writing-plans systematic-debugging test-driven-development verification code-review frontend-design project-analysis
"Use PostgreSQL to store users and SendGrid for emails"

Worktrees

The agent can isolate changes in separate git worktrees. Each feature or fix gets its own branch and working copy — no conflicts, no accidental overwrites. Clean separation between parallel workstreams.

Isolated Git Worktree

Frequently Asked Questions

What AI model does the agent use?

Teta Agent is powered by Claude by Anthropic. Sub-agents use lighter models for speed. Image generation uses Google Gemini.

Can the agent break my code?

In supervised mode, every file write, edit, and terminal command requires your explicit approval. You control what gets changed.

What are MCP servers?

Model Context Protocol servers are external tools the agent can connect to — image generation, browser automation, documentation lookup, database access. They extend the agent's capabilities.

How do sub-agents work?

The agent spawns lightweight sub-agents for parallel tasks like research, file exploration, and code review. Each works independently and reports back without cluttering the main conversation.

What are Superpowers skills?

Structured development workflows the agent follows — brainstorming, planning, TDD, debugging, code review. Each skill enforces senior-level engineering practices.

Is the agent free to use?

Yes. You get 3,000 free credits per month. No credit card required. Each credit is one AI message.

Try the agent.

3,000 free credits per month. No credit card required.

Get Started