Permission-bounded tool runs
Only approved tools run. Calls stay inside the scope you set for that AI employee—rejected at runtime when they step outside.
Built-in
Run approved tools, get a receipt every time
AgentX is the engine + common tool pack inside AgentOS X (not a separate product). AI employees use approved tools inside the scope you set; every tool run is attributable to a named AI employee. Each run returns a structured record—who ran which tool and what happened—not a chat transcript. You supply third-party credentials; AIAGENTNET does not bake your CRM or mail keys into the platform.
On the stack
Engine & tools on the stack
Built into AgentOS X (not a SKU): tools, bounds, receipts.
What this product owns
AI
AI employee
Day-to-day execution
running
- IdentityNamed AI
- ToolsConnectors
- PackagesInstall
- Agent WebOpen edge
Focus on this product
L3
Human
Judgment, approval, relationships—irreplaceable
L2
AgentOS X
Company OS & workspace you can self-host
L1
AI employees
Day-to-day work with identity and receipts
L0
Your data
Records under your control
Focus
Step 01 of 03
Execute
AI employees act with named ID inside OS bounds.
Step 02 of 03
Approved tools
Approved connectors fire under permission.
Step 03 of 03
Receipt
Every run leaves an audit trail toward data.
Replace, for the irreplaceable.
AgentOS X engine & tools
AgentOS X engine & tools
Run approved tools, get a receipt every time
AgentX is the engine + common tool pack inside AgentOS X (not a separate product). AI employees use approved tools inside the scope you set; every tool run is attributable to a named AI employee. Each run returns a structured record—who ran which tool and what happened—not a chat transcript. You supply third-party credentials; AIAGENTNET does not bake your CRM or mail keys into the platform.
Part of AgentOS X—not sold as AIAGENTNET X. Protocol details (identity, connectors, public API) live on the docs site. Credentials stay yours.Capabilities
What this product actually does
Named AI calls
You can see which AI employee ran a tool; your own product credentials stay separate at the third party.
Tool marketplace connectors
Discover and reuse tool-style connectors. Build to publish; Use to attach credentials and run—open ecosystem, not a closed plugin island.
Structured execution receipts
Every run leaves JSON evidence: agent identity, tool name, outcome. Audit and handoff use the receipt—not a free-form chat log.
How it works
From join to receipt
- 01
Bound the agent
Identity and permission scope are set before tools open.
- 02
Select connectors
Pick approved connectors; document which credentials your organization must provide.
- 03
Call with a named AI employee
The runtime shows which AI employee is calling; connectors can log that for safety.
- 04
Return a receipt
Pass/fail, tool list and agent id ship back for review and audit.
Who it is for
Buyers this page is written for
Teams wiring AI to real systems
CRM, mail, docs, internal APIs—without anonymous automation.
Builders listing tools
Publish connectors that accept named AI callers and grow the open market.
Operators who need audit
Every run must answer who called, which tool, what happened.
What you get
Concrete outcomes—not a feature laundry list
- Engine + tool pack delivered with AgentOS X (not a separate SKU)
- Named AI identity on every tool run
- Connector path for build and use
- JSON receipts suitable for audit and human review
- Developer docs (market-specific host)
Trust is part of the product. Audit and permissions, deployment boundary and data handling are built in. Read the trust brief or talk to sales.