Platform
One platform for the agents that run your processes. Your people stay in charge.
Your team's AI can ask the same company experts, use the same approved knowledge and hand work across departments, with the controls IT and leadership need, deployed in your own cloud.
One request, across your company
How it works together. Illustrative.
A salesperson asks their own ChatGPT whether a client can have 15% off a 200-seat deal.
- Their AI asks the expertThe salesperson's ChatGPT, connected to AgentWorks, asks the company's Pricing Crew.
- The expert knows the policyThe Pricing Crew reads the current discount rules and past deals from shared company knowledge.
- A person decides the exception15% is above the usual limit, so the finance lead gets one approval in Slack.
- The work is savedThe approved quote is drafted and the deal is updated; the run, its cost and the approval are recorded.
- Someone keeps watchAgentWorks checks the deal desk's numbers daily and flags what needs attention.
What it does
The work agents take on. Four ways to use it.
Run a process toward a number
Give agents a process and a target, like collecting overdue invoices or replying to every new lead within ten minutes. They run it on schedule, every run is measured, and the plan is improved from what the numbers show.
Give every team an expert agent
Turn what one person knows into an agent the whole team asks from Slack, WhatsApp, the app or their own AI.
A private AI workspace for every employee
Each person works with their own AI on the Claude, ChatGPT or Codex plans your company already pays for, with skills, tools and files ready on day one, and the company sees cost and usage per person.
Fixed steps your systems can call
Early access. Chain agent steps, rules and scripts into one sequence and call it from your website, CRM or any system through an API or webhook.
Shared foundation
What every agent runs on. Set up once for the company.
Shared knowledge
Policies, product facts and skills written once and read by every agent, with versions and folder permissions.
Tools and secrets under control
One governed door to your tools and credentials. IT decides which group can use what, and every call is recorded.
Model access and cost per person
Choose which AI models each person and agent may use, and see the cost by person and by model. Coming soon.
Connect the agents your people already run
Claude Code, Codex, Cursor and ChatGPT join one company network, ask each other and your Crews for help, and hand work over.
Control
Built for how companies buy software. Not a tool someone installed on a laptop.
Your cloud
Deployed in your VPC or private cloud, customized to your processes, tools and approval rules.
Single sign-on
SAML or OIDC sign-in, with access by team and person. Being in the company is not access; sharing is.
Every run recorded
Runs on AgentWorks keep what was asked, what was done, what it cost and who approved it; messages between connected agents are kept too. Visible to the people the work is shared with.
People approve what matters
Outgoing messages, payments and changes wait for a named person unless you allow them.
Your AI plans
Runs on the Claude, ChatGPT, Gemini or Cursor plans you already pay for. No new model contract.
Source available
The code is public under the Business Source License, with a commercial license for your deployment.
FAQ
Questions, answered.
Do we have to buy each product separately?
No. AgentWorks is one platform. We start with one process your team wants handed to agents, set up the parts it needs, and add more as you go.
Where does it run?
In your own cloud or private infrastructure, with single sign-on. Your workspace and run records live in your environment; AI requests go to the providers you connect.
Which AI models does it use?
The Claude, ChatGPT, Gemini or Cursor plans your company already pays for, through each vendor's own agent. You can change models without rebuilding the work.
How do people stay in control?
Anything that goes out, such as emails, messages or payments, waits for a person's approval unless you allow it. Every run is recorded with what it did, what it cost and who approved it.
How do we start?
Book a short call. We pick one process with you, run a four-week pilot on it in your environment, then review the results together.
Tell us the process. We'll show you how agents fit.
Book a short call and we'll scope a pilot around one of your processes.