Jelly: Local Multi-Bot Agent Workspace
Run persistent AI agents on your own Mac, Linux box or VM, with a real mobile web UI and a local Chrome you can take over.
- Source repo
- dctanner/jelly
- Stars
- ★ 67
- Last updated
- today
- License
- MIT
- Primary language
- TypeScript
- FA score
- 53/100 · Major gaps
At a glance
- Works with
- Platform-specificChatGPT · OpenAI API
- You'll need
- Typical use
- A solo founder running an entire business from one VM who wants separate agents per project directory to handle ops, scraping and routine tasks.
- Main limitation
- It is explicitly not a sandbox: agents execute shell commands and access files with your OS user's permissions and without per-tool approval, so anyone with network access can control its tools.
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-20+0.9 ★/day over 7 days
- Source review
- 53/100 · Major gaps
What does this agent do, and when should you use it?
Jelly is a local-first multi-agent system aimed at people running an entire business from a single VM. It is built with Bun, React, SQLite and the Pi agent harness; development serves on port 5173, production on port 3100, instance data lives in .jelly/ by default, and user-wide MCP configuration plus OAuth caches live in ~/.jelly/. You can create as many agents as you want, group them into project directories, and each agent gets its own browser window, profile and private desktop. Running agents requires a ChatGPT subscription or an OpenAI API key — there is no no-key demo mode — and model conversations are sent to your chosen provider. Jelly describes itself as a trusted-user tool rather than a sandbox: agents run shell commands and touch files with your OS user's permissions, while sudo requests go through a private authentication card whose password is never sent to the model or stored by Jelly.
Agents ship with shell tools, subagents and MCP connections, plus file uploads, inline images and downloadable artifacts. The structured browser tools let an agent inspect bounded visible text and element references, click and fill non-secret fields, manage session-local tabs, scroll, wait for text or readiness, and read redacted error diagnostics; these share the per-agent isolation and private-control gate, references expire on new observations, navigation or handoff, and snapshots cover the top-level DOM only. When an agent needs to sign in somewhere, you take over its browser via VNC inside the web UI; clipboard transfers are text-only, capped at 12,000 characters, and never reach agents, chat history or logs. The apply_patch tool adds Codex-style multi-file add/update/delete/move patches on both OpenAI API and ChatGPT, requiring exact unique context, rejecting symlinks and hardlinked sources, capping patches at 1 MiB per file/patch, 64 operations and 16 MiB combined working set, with non-transactional writes reported in a mutation ledger. Administrative work goes through request_sudo and needs Linux, sudo with askpass support and /usr/bin/python3. On the first message, Jelly makes one small tool-free OpenAI request to suggest a sea-themed agent name.
- A solo founder running an entire business from one VM who wants separate agents per project directory to handle ops, scraping and routine tasks.
- Someone who must let an agent sign into a third-party site but refuses to hand credentials to a model: the agent requests a login, then the human takes over via VNC in the web UI.
- A user with an existing ChatGPT subscription who wants to avoid buying separate API quota, connecting via the one-time device code in Settings.
- Sysadmin chores such as installing packages or managing services, where the agent calls request_sudo and the human approves through a private authentication card.
- Remote use from a phone or laptop: enable Tailscale and add the web UI to the iPhone home screen for near-native feel.
- Developers needing multi-file edits in one shot: apply_patch adds, updates, deletes and moves files over OpenAI API or ChatGPT.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Ports
5173·3100- Environment variables
FIRECRAWL_API_KEYJELLY_BROWSER_PATH- Config files
.env.local- Runtime
- Bun 1.3.9+ · Node.js 24+
- Install notes cover
- Linux
How do you install or deploy this agent?
Requires Bun 1.3.9+ and Node.js 24+; the managed browser desktop and sudo handoffs require Linux.
git clone https://github.com/dctanner/jelly.git
cd jelly
bun install
bun run devOpen http://127.0.0.1:5173. Local environment values go in .env.local, which is Git-ignored; set FIRECRAWL_API_KEY there to enable web search and fetching.
For production:
bun run build
bun startThen open http://127.0.0.1:3100, stopping development first since both modes use API port 3100 by default.
On Debian/Ubuntu x86-64 with Google Chrome or Chromium already installed, install the optional browser desktop:
bun run setup:desktopSet JELLY_BROWSER_PATH if the browser lives outside its usual system location. To reach Jelly over a private tailnet:
bun run dev:tailscaleHow do you use this agent?
Before sending a message you must connect an account or add a key in Settings — ChatGPT sign-in uses a one-time device code, so open the approval link shown in Settings and enter the code. Automatic mode prefers ChatGPT, then an API key; there is no no-key demo mode.
Click the top-left + to immediately create and open New Agent, which inherits the current project or gets a private workspace; before the first message, click the centered name or avatar to edit in place. Editing and saving the name before the first message opts out of automatic naming.
For eligible ChatGPT subscriptions (Pro 500 or Enterprise) or OpenAI API access, select GPT-6 Astra Ultrafast in Settings → Model or the composer's model menu; it requests Astra with service_tier: "ultrafast" while standard Astra stays the default, and account-access errors are surfaced rather than silently falling back.
In the Agent computer panel choose Take control to drive a given agent's browser, then use Paste to remote and Copy from remote for text clipboard transfers; if control is stuck after a session expires, use Agent computer → Recover lost control… to disconnect the old controller and clear private tabs and the remote clipboard while stored website sign-ins remain.
Development checks:
bun run check
bun run test:desktopbun run check covers TypeScript, structured-browser Chromium tests, patch filesystem tests and local provider-transport tests; bun run test:desktop additionally requires the Linux desktop runtime. Tests use temporary data and deterministic fixtures and need no paid model calls.
What are this agent's strengths and limitations?
- Per-agent isolation is concrete: separate browser windows, profiles, tabs, clipboard and human-control state, with at most eight browser sessions allocated at once.
- Credential handling is specific: sudo passwords are entered in a private card, never sent to the model or saved by Jelly, and clipboard transfers are text-only, capped at 12,000 characters and excluded from history and logs.
- You can use an existing ChatGPT subscription instead of buying separate API quota, and the device-code login works over Tailscale without a localhost redirect.
- apply_patch gives Codex-style multi-file moves with explicit size and operation limits plus a mutation ledger for partial failures, alongside the unchanged Pi edit/write tools.
- The local data layout is clear and auditable: .jelly/ holds conversations, credentials, browser profiles and artifacts while ~/.jelly/ holds MCP configuration and OAuth caches.
- It is explicitly not a sandbox: agents execute shell commands and access files with your OS user's permissions and without per-tool approval, so anyone with network access can control its tools.
- Model-provider lock-in: running agents requires a ChatGPT subscription or an OpenAI API key with no no-key demo mode, and local storage does not mean local inference.
- Platform limits are real: the managed browser desktop and sudo handoffs need Linux, and the desktop runtime now also requires xclip and /usr/bin/python3.
- apply_patch writes are not transactional, so a failure can leave partial changes, and paths are not sandboxed; symlinks and hardlinked sources are rejected.
- Setup is multi-step: Bun 1.3.9+ and Node.js 24+ are both required, and development and production share API port 3100, so you must stop one before starting the other.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Jelly: Local Multi-Bot Agent Workspace This agent | 53 · Major gaps | — | ★ 67 | today | TypeScript | ChatGPT · OpenAI API |
| Auto Browser | 76 · Some gaps | Self-hosted serviceFree + model costs | ★ 929 | 5d ago | Python | OpenAI API · Claude API |
| Aiden | 61 · Some gaps | CLIFree + model costs | ★ 909 | 2mo ago | TypeScript | ChatGPT · OpenAI API · Claude API |
| JoySafeter | 53 · Major gaps | Self-hosted serviceFree + model costs | ★ 319 | 7mo ago | Python | — |
How does FollowAgents rate this agent?
Why each dimension lost points
The README says agents can run shell commands and access files with the OS user's permissions, without per-tool approval. Connected Chrome shares the signed-in profile's cookies, and sudo may run immediately when system policy allows; least privilege and confirmation therefore score 1. The README gives concrete information about provider data flow, credential locations, notifications without message contents, private authentication, and same-host Apps cookie risks, supporting 2 for data-flow transparency and sensitive-data handling. Dependency versions and an override are listed, but the packet shows no dependency audit or vulnerability response evidence, so dependency security scores 1. Stop cannot reverse navigation, clicks, or website effects; patch writes are non-transactional and may be partial, while service restarts can interrupt agents, so external effects and rollback score 1. The repository and MIT contributor attribution are visible, but publisher identity, maintenance ownership, and update path are unclear; source attribution scores 1.
The README is broadly consistent about local operation, authentication, browser capabilities, and risks. Package scripts and test files indicate coverage across several feature areas, supporting 2 for self-consistency. Bun, Node, browser, sudo, and external model accounts are explicit dependencies, but no availability guarantees or offline alternatives are shown, so dependency availability scores 1. The docs describe user-facing states or limitations for model access, Claude sign-in, app embedding, and partial patch failures, supporting 2 for failure messages. This static review did not run the code.
The intended user, single-VM workspace, project grouping, and desktop/mobile access scenarios are clear, but guidance for varied skill levels and deployment contexts is limited, so audience and scenarios score 2. The docs clearly state that Jelly is not a sandbox, network-accessible users may control tools, and Chrome shares accounts without an allowed-sites list, supporting 2 for capability boundaries. The supplied SKILL targets CSS transition implementation and has detailed trigger conditions and references, but it is not repository-wide Agent guidance; applicability triggers for the Agent product are less developed, so trigger precision scores 1. Setup instructions cover Bun/Node, macOS/Linux, Tailscale, Chrome, and authentication differences, supporting 2 for environment fit.
The README is organized by features, setup, security, service management, and development, with links to focused docs, so information architecture scores 2. Quick start gives Bun/Node requirements, commands, authentication prerequisites, and optional environment configuration, though installation details are spread across guides, so install notes score 2. Package, repository, and Agent terminology are mostly stable, supporting 2 for naming stability. The README includes examples and linked guides, but the packet does not show a FAQ or broad troubleshooting Q&A, so examples and FAQ score 1. It documents the non-sandbox model, shared cookies, non-transactional patching, notification and browser restrictions, and restart effects in detail, earning 3 for known limitations. The MIT license text is complete, earning 3. package.json shows version 0.1.0 and scripts, but the packet contains no changelog or release compatibility policy, so versioning and changelog score 1. Contributor attribution exists, but no individual or organization is named as maintainer, so maintenance responsibility scores 1.
The product offers multi-agent workspaces, persistent sessions, browser collaboration, subagents, tools, and notifications. The test inventory shows separate coverage in several areas, but this static review cannot confirm execution, so output usability scores 2. Project directories, persistence, browser handoff, and subagents provide useful capabilities beyond a basic chat interface, supporting 2 for marginal value. Benefits depend on the user managing a high-privilege local environment, model access, and browser accounts; the packet provides no resource or deployment cost assessment, so cost-benefit scores 1.
Key README limitations and operational claims link to focused documentation, and the server entry point and package configuration are available for inspection, supporting 2 for claim traceability. The README, package scripts, server entry point, and test samples corroborate local startup, model configuration, and some behaviors, supporting 2 for cross-source corroboration; not every documented claim has implementation or test evidence in the packet. The README distinguishes demo data from real capabilities and says sudo and website sign-in in the launch video are simulated; test samples also identify isolated fixtures or stubbed model behavior, supporting 2 for fact-inference separation.
- This is a high-privilege local tool, not a sandbox; network-reachable users may control its tools. Keep it on loopback or a trusted Tailscale network, and configure model access, browser accounts, and sudo carefully.
- Connected Chrome shares cookies and accounts in the selected Chrome profile. Browser actions, website effects, and non-transactional patch writes may not be recoverable through Stop or rollback.
- This assessment is a static review of the supplied file sample; no code or tests were executed, and unseen files are unknown rather than missing.
FAQ
Does Jelly cost anything to run?
Can agents run commands as root?
Can I expose it to the public internet?
Does Stop undo what an agent already did?
Can I run it on macOS?
Is Jelly: Local Multi-Bot Agent Workspace open source?
Is Jelly: Local Multi-Bot Agent Workspace safe to use?
Is Jelly: Local Multi-Bot Agent Workspace still maintained?
Who makes Jelly: Local Multi-Bot Agent Workspace?
Thanks — we use this to improve the page.