Errand
runerrand.dev- Category
- AI Agents
- Rank
- No. 1270Tools index
Previous survey · No. 1241 ·
- Pricing
- Open Source
- Type
- AGENT
- Use case
- Agent Building
- Interfaces
- Desktop
- GitHub
- 140 stars
- Latest release
- v0.2.6
- Date
About
Errand is a macOS app for working with persistent AI agents, each given its own cloud computer (via Runta) with files, browsers, and tools scoped to that agent's workspace. Conversations and workspace state persist between sessions, so an agent can be reopened later to continue exactly where it left off, and the project is open source as a reference application on top of Runta Cloud Agents.
At a glance
Errand is a Mac client for persistent AI teammates, each backed by a Runta cloud computer. Consider it when a task benefits from keeping its conversation, files and working environment between sessions. That persistence also brings hosted-data handling and usage charges; the public browser preview demonstrates the layout, not an agent completing your task.
Good fit
People who want to return to the same agent workspace for ongoing research or development rather than starting a disposable conversation each time. The client presents conversation, activity, artifacts and the cloud desktop together. Its source routes messages and artifacts through Runta’s agent API, so the persistent workspace is a remote service rather than a computer running locally inside the Mac app.
Consider something else when
You need offline work, a local-only assistant or proof of task quality before creating a cloud account. The website’s sample checkout investigation is canned. Entering a new synthetic task returned an explicit notice that the preview cannot run it, so it can help assess navigation but cannot demonstrate reasoning, browser work or successful completion.
Start here
Explore the public preview first: choose a teammate and open its sample artifact to judge whether the conversation-plus-output arrangement fits your work. Actual use requires the macOS app, a Runta account and configured model access. Begin there with a disposable task, then inspect the output and cloud resource state before entrusting it with ongoing work.
Cost and requirements
The desktop app targets Apple Silicon Macs on macOS 13 or later. Runta’s checked rates are $0.0504 per vCPU-hour, $0.0162 per GiB-hour of memory and $0.000108 per GiB-hour of disk, with model access separate. Running charges all three; paused still charges memory and disk, and suspended/shut-down resources still charge disk until deleted. Closing the client is not that deletion action.
Limits to know
Runta’s privacy notice covers runtime files, commands, credentials, checkpoints and logs as customer content; this is a consequential hosted-data choice. Optional Token Analyze tool-I/O capture is off by default and has a stated seven-day retention when enabled, not a general seven-day deletion rule for all content. Only sample-preview navigation was tested; no real agent, desktop installation or output-quality assessment was performed.
Sources checked 2026-10-30. Source-based assessment.
- runerrand.dev
- github.com/runta-dev/errand/blob/main/README.md
- github.com/runta-dev/errand/blob/main/src/clients/http/RuntaCloudAgentsClient.ts
- github.com/runta-dev/errand/blob/main/electron/preload/index.ts
- github.com/runta-dev/errand/blob/main/src/shared/runtaEndpoints.ts
- runta.com/pricing/
- runta.com/privacy-policy/
Why it made the leaderboard
Gives each AI agent its own persistent cloud computer and conversation, so work like research, drafting, or bug fixes continues across sessions instead of resetting each time.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.