Mochi

Mochi builds, runs and looks after friendly AI agents that plan, use your tools and finish real work — while you stay in control.

Introduction

Mochi turns a plain-language goal into an AI agent that plans its own steps, uses your tools and reports back — no code required. To get one running, you need just four things.

01

A clear goal

The outcome in plain language: what to do, how often, and what “done” looks like.

02

Your tools

The apps and data it should work with — email, docs, databases or your own API.

03

Guardrails

Permissions, spending limits and the actions that must wait for your approval.

04

A place to run

Mochi’s cloud runtime runs it on a schedule or a trigger. Nothing to host.

How it works

Six steps, no code. You decide what the agent should achieve and what it may touch — Mochi handles the planning, running and reporting.

Goal

Summarise new tickets daily

Run
01

Describe the goal

Write what you want done in plain language. Mochi turns it into a clear brief with success criteria.

  • Prompt
  • Brief
  • Success criteria
02

Connect your tools

Link the apps, files and APIs the agent may use — each with scoped, revocable access.

  • Apps
  • Data
  • APIs
Read inboxAuto
Send repliesAsk me
Spend over $50Ask me
03

Set guardrails

Choose what it can do on its own, what needs your approval, and how much it may spend.

  • Permissions
  • Limits
  • Approvals
  1. 1Fetch new tickets
  2. 2Group by topic
  3. 3Draft summary
  4. 4Post to #support
04

It plans

The agent breaks the goal into steps and picks the right tool for each one.

  • Reasoning
  • Steps
  • Tool choice
›zendesk.search(status:new)
✓42 tickets fetched
›cluster(topics)
✓6 topics found
›slack.post(#support)
05

It acts

Steps run in a sandboxed cloud runtime, on a schedule or a trigger, with every action logged.

  • Runtime
  • Logs
  • Retries
06

You review

Read the run report, approve or correct — and the agent remembers your feedback next time.

  • Reports
  • Feedback
  • Memory
Agent templates · 04

Start from a ready-made agent or build your own. Hover to peek, click to see exactly how each one runs.

The agent loop

Every run follows the same four-step loop. Keep scrolling to follow one through it.

01 —

Plan

Reads the goal, the context and what happened on past runs, then writes a step-by-step plan.

  • Goal
  • Context
  • Memory
02 —

Act

Calls your tools — searches, writes, sends, updates — one verified step at a time.

  • Tools
  • APIs
  • Sandbox
03 —

Observe

Checks every result against the plan and your guardrails, and pauses for approval when needed.

  • Checks
  • Guardrails
  • Approvals
04 —

Learn

Summarises the run, keeps what worked and gets sharper with every piece of feedback.

  • Reports
  • Feedback
  • Memory
Get started

Tell us what you'd like to automate — we'll help you launch your first agent.

I want to automate

Talk to us

Early access is opening in waves.
We reply within two working days.
Mochi
© 2026 Mochi. Example agents shown for demonstration.
AI agent platformWAKING AGENTS
Mochi
Plan · Act · Learn000