Skip to content
Zod Agent

Make it yours. A client-side agentic AI chat workspace that runs entirely in your browser.

Zod Agent gives you a hub-first AI workspace on top of any OpenAI-compatible API — with a real tool loop, structured responses, YAML configuration, agent workflows, and scheduled background agents. No account, no server, no data leaving your machine.

Your entire workspace — sessions, memory, tools, provider keys, and preferences — lives in your browser. Close the tab and come back later: it's all still there. The only requests leaving your machine are the ones you send to the providers you configure.

One workspace, every feature.

Hover to replay the full UI/UX demo — streaming chat, the tool loop, confirmation plans, agent runs, code, search, scheduling, orchestration, and more, all in one loop.

Built for real work.

Zod is a general-purpose agent workspace — here's what people use it for.

Research & market scans

Point a Research agent at a topic and get findings, sources, and summaries back in a structured card — scheduled daily so the report is waiting for you.

See the agent run

Coding partner

A Coder agent that reads your memory of the project, proposes a plan, asks before it changes anything, then applies the work and shows you a diff.

Review the flow

Scheduled intelligence

Digests, stand-ups, and status reports on a cron — once, interval, daily, or weekly. A service worker keeps them running even when the tab is closed.

Try the scheduler

Teaching & demos

Show your audience a live agent loop with visible tool calls and reasoning — no accounts, no hidden backend, just the browser.

Watch it stream

Desktop utility

On desktop, Zod can reach into apps, files, and notifications — with a native shell around the same frontend you use on the web.

See the desktop app

Private & local-first

Sessions, memory, and keys live in your browser. Only the requests you choose go to providers you configure — nothing is ever sent to a Zod backend (there isn't one).

Read the security notes

Zero backend. Provider-agnostic.

Point Zod at OpenAI, OpenRouter, Together, Groq, or a local server like Ollama or LM Studio — anything that speaks /v1/chat/completions. Configure many providers, pick a default and a fallback.

Configure providers

Tool-using agents in the browser

A real multi-round function-calling loop runs client-side. The model can call 9 built-in tools, your own custom JavaScript tools, and MCP server tools over HTTP — and keep going until the task is done.

Explore the tool system

Structured, rendered responses

Zod parses lightweight section markers into rich cards — thinking steps, badges, findings, todos, code blocks, files, errors, security warnings, visualizations, and Approve/Reject confirmation plans.

See how it renders

Workflows that actually run

Compose agent workflows in the orchestrator — steps, tool grants, skills, and @goto transitions — and execute them against your provider with live step-by-step progress.

Build a workflow

Scheduled & background agents

Schedule prompts once, on an interval, daily, or weekly. A service worker keeps running scheduled agents even after you close the tab — and the results land in your sessions when you return.

Learn about scheduling

A toolbelt in every conversation

Every reply comes with real tools the agent can call — not just text.

calculator

Safe arithmetic with a hand-written parser.

todo

A persistent to-do list that survives sessions.

memory

Long-term markdown memory blocks.

clipboard

System clipboard read and write.

timer

In-browser reminders with notifications.

code_tools

JSON & YAML utilities.

confirm

Ask before destructive actions.

random

UUIDs, dice, numbers, and hashing.

time

Timezones and conversions.

+ custom tools & MCP

Make it yours.

No account. No server. Add an API key and your agent is ready — in your browser, as a Chrome extension, or as a desktop app.