01
Your CLI agents, first-class
Claude Code, Codex, Antigravity, Copilot, and Cursor run directly on your local runners with real access to your repositories, files, and tools — not a sandboxed clone.
For power users & developersRun your own CLI agents on your machines. Any model.
Claude Code, Codex, Gemini, and more — running on your own Mac, Windows, or Linux box with real repo and file access. Connect in minutes. No VPN, no open ports.
- Your CLI agents, your machines, your models
- Connect in minutes — no VPN, no open ports
- Real repo, file, and CLI tool access
Local runnerConnect any machine in minutes
j5agentfleet.com/dashboard
Any modelClaude · GPT · Gemini · local
Built for people who already drive AI
The orchestration layer for your agents.
You already use Claude Code and Codex. J5 Agent Fleet runs them as first-class agents on your own hardware, coordinates them across real work, and lets you pick the right model for each job.
02
Connect any machine, fast
Register any Mac, Windows, or Linux box as an execution node — the desktop app for Mac and Windows, or a single CLI command for Linux and headless. Outbound poll only: no inbound ports, no VPN, no exposing your network.
03
Bring any model
Pin the right model per agent — Anthropic, OpenAI, Bedrock, or a local model via LM Studio. Mix API and CLI-backed models across one project.
04
Convene agents in a meeting
Put multiple agents — each with its own model, runner, and tools — around one topic, seeded with your repos and docs. They take turns, build on each other, and stop when they align or hit your budget cap.
05
Local browser & computer control
Optionally enable agents to drive authenticated browser profiles and (on macOS) the desktop — approval-gated and turned on per runner, with actions recorded to the control plane.
Agents that act in your stack
- Gmail
- Slack
- GitHub
- Notion
- Linear
- Google Drive
- +250 tools
Ship with it
Point your agents at real work.
Connect a machine, pick your models, and start delegating. The free tier runs real tasks without a credit card.
Start freeNo credit card required for the free tier.
FAQ
Developer questions, answered.
How local runners, CLI agents, models, and control work when you put J5 on your own machines.
Yes. The agent runner daemon lets you connect any Mac, Windows, or Linux machine as a local execution node. Mac and Windows users can download the desktop app from Settings → Runners and sign in; the app registers the daemon automatically. Linux, headless, CI, and advanced manual installs can still use one-time setup tokens and CLI commands. Agents then run with full access to your local repositories, CLI tools, and environment.
A single assistant can help with one conversation at a time. J5 Agent Fleet is designed for coordinated execution across multiple agents, tasks, and workflows, with project structure, context management, reusable task patterns, and operational visibility built in.
Yes. Teams can expand the fleet by creating their own specialists with a role, category, model, and operating instructions that fit the work they actually need to run.
Optionally, yes. For projects with a valid local workspace, J5 can maintain project-scoped QMD knowledge and show configuration status directly in the product. It is an operational aid for grounded project context, not a claim of universal memory.
Approval gates are human checkpoints that block agent execution until a team member explicitly signs off. They are useful before high-risk actions, external-facing changes, spend-heavy tasks, or any step where a human decision is genuinely required. Gates can be configured by checkpoint type, risk level, or task type, and the platform keeps a full audit trail of who approved what and when.
J5 Agent Fleet offers a free Trial to get started, a Free tier with a monthly task allowance, a Pro tier for individual power users, and a Team tier for collaborative workspaces. See the Pricing page for current limits. Paid plans are billed through Stripe and can be managed from Settings → Billing at any time.
Yes. The Free tier includes a monthly task allowance so you can run real work without a credit card. When you reach the limit you can upgrade to Pro or Team from Settings → Billing. Trial accounts also get an extended window to explore the full feature set before choosing a plan.