The platform your AI team runs on.
Genfleet runs every agent in isolation, connects it to the systems your business already uses, keeps a record of its work, and routes each request to the right model.
Genfleet
Isolated by default.
Agents run code on your behalf, so each one is boxed in: its own sandbox, inside a workspace that belongs only to your company.
Your workspace
Support
Own sandbox
Finance
Own sandbox
Sales
Own sandbox
Another company
A sandbox for every agent
Agent code never shares machinery. Each agent runs locked down, with its own limits.
A workspace for every company
Agents, data, and settings are scoped to your company and invisible to anyone else.
Secrets kept out of agent code
API keys and connected accounts live in a secrets store, and every read is logged.
Outbound traffic in view
See every outside address an agent reaches, so nothing talks to the internet unnoticed.
Connected to the systems you run on.
A connector brings a system’s tools to your agents, like reading a Slack channel, updating a Jira issue, or looking up an order in your ERP.
- Slack
- Jira
- Linear
- GitHub
- ERP
- CRM
- Calendar
Connect once
Your company signs in to each system once. Every agent that needs it can use that connection.
Disconnect anytime
Remove a connection whenever you like, from the same place you added it.
Custom connectors
Running a system nobody else does? We build the connector with you during onboarding.
See what every agent did.
Agents work while your team is busy. Genfleet keeps the record, so anyone can open a run later and follow it step by step.
Today
Where is order 4821?
Support agent, 09:14, finished in 2.4s
A log of every run
Which agent ran, what it was asked, and how long it took.
Every step visible
Open any run to see each tool call, in order.
Stats for every agent
Each agent’s page shows how it has been working, next to its tools, memory, and logs.
People stay in the loop
Hold sensitive actions until someone on your team signs off.
The right model for every task.
Your agents should not care which AI provider answers them. The Genfleet model router sits in between, so you can change models without changing agents.
A request comes in
It goes to
- Fast modelQuick answers, lowest cost
- Balanced modelEveryday work
- Reasoning modelHard, multi-step problems
A short, simple question, so the fast model answers it in a moment for a fraction of the cost.
Your keys or ours
Use your company’s own model provider keys on every plan, or let us manage them for an extra charge.
Automatic fallback
If a provider slows down or goes offline, requests move to a healthy one.
Budgets and usage per agent
Set spending limits and see what each agent costs, by model.
Build your own agents.
Create a custom agent right in your workspace, or bring one your team has already built. Either way it runs with its own sandbox, connectors, oversight, and the model router.
Build it in your workspace
No code needed. Give the agent a role, choose its model, and add the tools and company knowledge it should use.
New agent
Returns assistant
- Role
- Answers questions about orders and returns
- Model
- Balanced model
- Tools
- ERP, Email
- Knowledge
- Return policy, Shipping FAQ
Convert one you already have
Add the genfleet-sdk skill to your favorite coding agent, like Claude Code, Cursor, or Codex, and ask it to convert your agent. Then push it to Genfleet for review.
1. In your terminal
$ npx skills add genfleet/sdk2. Then ask your agent
› Convert my support agent to the Genfleet SDK
3. Sign in and submit it for review
$ gf login
$ gf agents push ./support-agent
See the platform with your own work.
We will walk you through isolation, connectors, and oversight using the processes your team actually runs.