Jev needs Jeeves. Open source productivity workflows with Jev decisions, AI agents, and a place for your input.
See the codePut Jev to work in everyday workflows.
Jeeves is an open source app for productivity workflows. Jev, TypeSafe's decision model, chooses what happens next. Agents handle the writing, research, and browser work. Jeeves connects the steps and waits when it needs your answer.
For example, Jev can decide whether a request describes an outage, a small fix, or something that needs more detail. Jeeves sends it to the right agent or asks you a question. You can inspect the decision and reuse the workflow with the next request.
Download Jeeves · Quick start · Watch the demo · Website
You know the routine: paste the notes, explain the task, ask for a draft, check the draft, remember what worked. Then do it all again next Tuesday.
Jeeves lets you keep that routine as a workflow you can see and change.
No Node.js, terminal, or Jeeves account needed. Choose your system from the preview.11 release:
| Your computer | Download | Install |
|---|---|---|
| Mac, Apple Silicon (M-series) | Apple Silicon DMG | Open the DMG, drag Jeeves to Applications |
| Mac, Intel | Intel DMG | Open the DMG, drag Jeeves to Applications |
| Windows, x64 | Windows installer | Run the installer, then open Jeeves |
| Linux, Debian/Ubuntu x64 | Debian package | Install with your package manager |
| Linux, x64 portable | AppImage | Make executable, then open |
This is an early preview. Mac downloads are signed and notarized by Apple. Windows installers are unsigned and may show a security warning. Installation help, checksums, and first-run instructions.
The example needs no API keys. Its output is clearly labeled as a sample. For an AI answer to your own notes, choose Use my own notes, connect one agent service, and switch to Live.
Open Set up my AI connections during onboarding, or Home → Set up AI later. Save and verify the services you want to use.
| What you want to do | What you need |
|---|---|
| Try the sample or edit a workflow | Nothing else |
| Write or analyze with agents | One of OpenAI, Codex CLI, OpenRouter, or a local model server |
| Add Jev decisions | A TypeSafe API key, plus an agent service if the workflow has agent steps |
| Run browser tasks | Google Chrome and the live provider selected for that browser step |
Jev is TypeSafe's decision model. Jeeves is this app. A small difference in spelling; a useful division of labour. Hosted services have their own access requirements and costs. API credits are not included.
| Start with | What happens |
|---|---|
| Notes to weekly update | Turn project notes into a draft, check it against the notes, and review the result |
| Request triage · Jev | Route urgent and routine requests to different agents; ask you for detail when the request is unclear |
| GitHub maintenance | Read issues and pull requests, then prepare a maintainer brief |
| Grocery cart preparation | Collect your list, budget, and preferences before opening the browser; pause for login or review |
Find these in Explore or Templates. Each live workflow shows its required connections before you run it. Browser shopping stops for review before checkout.
Requires Git, Node.js 22.12 or newer, and npm. Quit the desktop app first so the local engine port is free.
git clone https://github.com/priyankark/jeeves.git
cd jeeves
npm ci
npm run dev
Open http://127.0.0.1:5173 and choose Try the example. No .env file is needed for the sample. Press Ctrl+C in the terminal to stop both development servers.
For a desktop window, stop the development servers and run npm run desktop. For a production web build, run npm run build followed by npm start, then open http://127.0.0.1:4317. More setup options.
Workflows, settings, and history live on your computer. Cloud AI services receive the task context you send them. Saved API keys are kept in a private local settings file, not in workflow exports or browser storage. The file is not encrypted. Website sign-ins use a separate local Chrome profile.
This preview has manual updates. Schedules need the local engine to stay running. Websites can change or ask you to sign in, and generated results still need your review. See the user guide for execution details and limits.
The release passed 136 engine and integration tests, 56 browser tests, and packaged-app checks on Mac, Windows, and Linux. These are automated checks, not a claim that every real-world task succeeds. Verification notes · Release checks
Found a rough edge? Report it. Got a useful routine? Contribute a workflow. Want to improve the app? Start with the contributor guide.
Quick start · User guide · Release guide · Security · Apache 2.0 license
“I will give the matter my best consideration, sir.”
Jeeves, in P. G. Wodehouse’s The Inimitable Jeeves (1923).
A fine brief for a software project. Literary inspiration, with no affiliation or endorsement implied. The app and first-party starter workflows use Apache 2.0; dependencies, models, and imported skills retain their own terms.
Try the example, then tell us how it went. We want to hear where you got stuck as well as what worked. The first ten minutes guide gives you a few things to try.
Launch materials: current launch kit. Older marketing exports are archived and should not be used for this launch.
TypeScript
74.4%
JavaScript
10.1%
CSS
10.0%
HTML
5.6%
Jev needs Jeeves. Open source productivity workflows with Jev decisions, AI agents, and a place for your input.
See the codePut Jev to work in everyday workflows.
Jeeves is an open source app for productivity workflows. Jev, TypeSafe's decision model, chooses what happens next. Agents handle the writing, research, and browser work. Jeeves connects the steps and waits when it needs your answer.
For example, Jev can decide whether a request describes an outage, a small fix, or something that needs more detail. Jeeves sends it to the right agent or asks you a question. You can inspect the decision and reuse the workflow with the next request.
Download Jeeves · Quick start · Watch the demo · Website
You know the routine: paste the notes, explain the task, ask for a draft, check the draft, remember what worked. Then do it all again next Tuesday.
Jeeves lets you keep that routine as a workflow you can see and change.
No Node.js, terminal, or Jeeves account needed. Choose your system from the preview.11 release:
| Your computer | Download | Install |
|---|---|---|
| Mac, Apple Silicon (M-series) | Apple Silicon DMG | Open the DMG, drag Jeeves to Applications |
| Mac, Intel | Intel DMG | Open the DMG, drag Jeeves to Applications |
| Windows, x64 | Windows installer | Run the installer, then open Jeeves |
| Linux, Debian/Ubuntu x64 | Debian package | Install with your package manager |
| Linux, x64 portable | AppImage | Make executable, then open |
This is an early preview. Mac downloads are signed and notarized by Apple. Windows installers are unsigned and may show a security warning. Installation help, checksums, and first-run instructions.
The example needs no API keys. Its output is clearly labeled as a sample. For an AI answer to your own notes, choose Use my own notes, connect one agent service, and switch to Live.
Open Set up my AI connections during onboarding, or Home → Set up AI later. Save and verify the services you want to use.
| What you want to do | What you need |
|---|---|
| Try the sample or edit a workflow | Nothing else |
| Write or analyze with agents | One of OpenAI, Codex CLI, OpenRouter, or a local model server |
| Add Jev decisions | A TypeSafe API key, plus an agent service if the workflow has agent steps |
| Run browser tasks | Google Chrome and the live provider selected for that browser step |
Jev is TypeSafe's decision model. Jeeves is this app. A small difference in spelling; a useful division of labour. Hosted services have their own access requirements and costs. API credits are not included.
| Start with | What happens |
|---|---|
| Notes to weekly update | Turn project notes into a draft, check it against the notes, and review the result |
| Request triage · Jev | Route urgent and routine requests to different agents; ask you for detail when the request is unclear |
| GitHub maintenance | Read issues and pull requests, then prepare a maintainer brief |
| Grocery cart preparation | Collect your list, budget, and preferences before opening the browser; pause for login or review |
Find these in Explore or Templates. Each live workflow shows its required connections before you run it. Browser shopping stops for review before checkout.
Requires Git, Node.js 22.12 or newer, and npm. Quit the desktop app first so the local engine port is free.
git clone https://github.com/priyankark/jeeves.git
cd jeeves
npm ci
npm run dev
Open http://127.0.0.1:5173 and choose Try the example. No .env file is needed for the sample. Press Ctrl+C in the terminal to stop both development servers.
For a desktop window, stop the development servers and run npm run desktop. For a production web build, run npm run build followed by npm start, then open http://127.0.0.1:4317. More setup options.
Workflows, settings, and history live on your computer. Cloud AI services receive the task context you send them. Saved API keys are kept in a private local settings file, not in workflow exports or browser storage. The file is not encrypted. Website sign-ins use a separate local Chrome profile.
This preview has manual updates. Schedules need the local engine to stay running. Websites can change or ask you to sign in, and generated results still need your review. See the user guide for execution details and limits.
The release passed 136 engine and integration tests, 56 browser tests, and packaged-app checks on Mac, Windows, and Linux. These are automated checks, not a claim that every real-world task succeeds. Verification notes · Release checks
Found a rough edge? Report it. Got a useful routine? Contribute a workflow. Want to improve the app? Start with the contributor guide.
Quick start · User guide · Release guide · Security · Apache 2.0 license
“I will give the matter my best consideration, sir.”
Jeeves, in P. G. Wodehouse’s The Inimitable Jeeves (1923).
A fine brief for a software project. Literary inspiration, with no affiliation or endorsement implied. The app and first-party starter workflows use Apache 2.0; dependencies, models, and imported skills retain their own terms.
Try the example, then tell us how it went. We want to hear where you got stuck as well as what worked. The first ten minutes guide gives you a few things to try.
Launch materials: current launch kit. Older marketing exports are archived and should not be used for this launch.
TypeScript
74.4%
JavaScript
10.1%
CSS
10.0%
HTML
5.6%