Skip to content
1Forgebot
Get started

Inside the app.

Each bot has a role, its own cloud browser, rules you set, and a memory you can read. It does the job and stops for the things that need a person: approvals, sign-in codes, and judgment calls.

Approval cardScreenshot from the live app
An approval card in a conversation with the bot Scout. It asks to send an email with the subject Toronto plumbers offering drain camera inspections, shows the recipient, body, and subject, and offers Approve, Approve and remember, and Deny. Below it, the run reads Waiting for your approval.
Scout asks before it emails a research summary. The card shows the recipient, subject, and body, and expires if no one answers.

Approval cards show the exact action.

When a bot wants to send, post, change a record, delete, share, or spend, the run pauses and a card appears in the conversation, in the approvals inbox, and as a phone notification.

The whole action, not a summary
The card lists the tool, the target, and every field: the recipient, the subject, the full body.
Bound to what you saw
Approval covers that exact action once. If the bot changes a word or a recipient, it has to ask again.
Edit, then approve
Fix the wording yourself. Only your edited version goes out.
Approve and remember
Let the same kind of action through for a limited time, so routine work stops asking.
Cards expire
A request left alone expires instead of firing days later. The card shows the time left.
Deny, and the bot adjusts
The action does not run, and the bot tells you what it can do instead.

Set rules per bot.

Every kind of action has a setting: do it without asking, ask me first, or hand it to me. A bot can follow the workspace or have its own, except where a higher level has locked a rule.

Bot settings, Rules tabScreenshot from the live app
The Rules tab for the bot Scout. Each kind of action, such as Read information, Write drafts, Change records in your apps, and Send emails, texts, and posts, shows its current setting and a menu to follow the workspace setting or override it.
Scout's Rules tab. Reading and drafting run on their own; sends and record changes ask first and get a second check.
A second check on risky actions
Sends, record changes, and deletes are reviewed by a separate check before they reach you, and anything that looks off is flagged.
Locked means locked
Password and security changes are always handed to a person, whatever a bot's rules say.

Every run leaves a timeline.

Open any task to see each step the bot took, in order, with the time it took, what it read, and whether it went through a connected app or its browser.

Run timelineScreenshot from the live app
A run page titled Scout's run. A summary shows Done, 3 steps, 21 seconds, 2 cents, and no browser time. Below, a timeline lists Run started, Model call, and several Web search steps with their results and links.
The run behind the recording on the home page: 3 steps, 21 seconds, 2 cents, no browser time.
Cost and time on every run
Each run shows its duration, model cost, and browser minutes, so you can see what a job really costs.
Proof of what happened
Searches, pages read, and tool results stay on the run, linked from the conversation.

A memory you can read and edit.

Bots keep what they learn as plain statements: facts, preferences, decisions, and corrections. Each one shows where it came from and who it applies to.

Bot settings, Memory tabScreenshot from the live app
The Memory tab titled What Scout knows. A list of memories, each labelled as a fact, preference, or decision, shows its source, such as from the bot's own review or a finished task, with an edit and delete menu. Filters for source, status, and scope sit above the list.
What Scout knows, including a lesson from the recorded run: check the site states free before calling an inspection free.
Corrections stick
Correct a bot in a conversation and it saves the correction for every run after.
Only this bot, or all bots here
Keep a note to one bot, or share it with every bot in the workspace.
Delete by source
Disconnecting an app stops new access. Remove everything learned from it in one step.
Your documents too
Add documents to the workspace knowledge library and bots can draw on them in their work.

Its own browser. You can take over.

Bots work in a real browser in the cloud, so they can use sites that have no API: supplier portals, booking pages, back offices. You can watch any of it live.

Illustration: you have the wheelLive view
Example: a sign-in asks for a code sent to your phone. You have taken control of the bot's browser and are typing the code; the bot is paused until you hand back.
Sign-ins without sharing passwords
Save a sign-in to the workspace vault. The bot's browser is filled in for it, and the model never sees the password.
Take control, then hand back
When a site asks for a code or a verification step, the bot hands it to you. While you have control, the bot doesn't type or click.
A fresh look after handback
When you hand back, the bot looks at the page again and tells you what changed before it continues.
Runs on our servers
Close your laptop and the job keeps going until it needs you.

Write to it like you would to staff.

Plain language from the web, the installable phone app, or the desktop app. A conversation picks up where you left it on any of them.

One conversation per bot
The history and context stay with the work.
Group conversations
Put several bots and people in one conversation. Mention a bot to bring it in.
Notifications that matter
You hear about approvals, questions, completed jobs, and anything that failed.
Desktop and phone
Get the apps from the download page, or install from your browser.

17 skills, on request or on a routine.

Skills are repeatable jobs a bot knows how to do. Routines start them for you.

  • Everyday work

    Daily briefing, Inbox triage, Research brief, Document summary, Weekly report, Data transfer

  • Sales

    Lead capture, Lead follow-up

  • Marketing

    Review responder, Social post drafts

  • Support

    Support reply drafts

  • Bookkeeping

    Invoice from job, Receivables follow-up, Bill intake

  • Front office

    Appointment booking, Appointment reminders

  • Calls

    RealVoice call follow-up

  • On a schedule

    Every weekday at 8 am, or the first Monday of the month.

  • When something happens

    A new email, a form submission, or a call that just ended.

  • From a webhook

    Another system calls the routine's private URL.

Teams of bots with a coordinator.

Give a group of bots a shared conversation. The coordinator breaks the job into tasks and assigns them, specialists report back, and every task has exactly one owner.

Ready-made teams
Install Front Office (coordinator, receptionist, inbox assistant) or Marketing Team (coordinator, research, writing, reporting).
Out in the open
Bots message each other inside the conversation, so you can read every handoff.
A shared task list
See who owns what, what is blocked, and what is done.
Your copy to change
A template installs as your own copy. Change any bot, skill, or rule afterwards.

Pick the model for each workspace.

Choose the model provider that suits the work and the budget, workspace by workspace.

  • OpenAI

    Connect OpenAI models for chat, drafting, and reasoning.

  • OpenRouter

    Reach a wide range of models through one OpenRouter connection.

  • DeepSeek

    Use DeepSeek models where they fit the work and the budget.

Workspaces keep everything apart.

A workspace holds its own bots, memory, files, browsers, sign-ins, connections, rules, and spending limits. Agencies on the Operator plan run one workspace per client.

Read how isolation, approvals, and the vault work on the security page, or see which plan fits on pricing.

Your first job ticket1Forgebot

Put your first bot to work today.

Pick a plan, name your first bot, and hand it a job. Prices are in Canadian dollars, billed monthly.