Connection and control layer

Define one trusted job.
Run it from every AI interface.

Onterix is the connection and control layer between AI assistants, agents, applications, and the systems your company runs. Define a renewal, vendor review, or other recurring job once; every run checks current access, controls what AI receives and changes, and records the result.

Connect your first system See the runtime Read the quickstart

Same capability, separate runs
Who starts the run	Interface	Trigger
Maya Chen · person	Codex MCP	On demand
Renewal monitor · agent	Onterix API	Schedule
Revenue service · app	Onterix SDK	Workflow event

Onterix runtime controls

IdentityData releaseApprovalSafe executionEvidence

ONTERIX RUNTIMEControls apply to each run

Current source accessReleased dataApproved changesConfirmed result

Systems and outputs

KnowledgeBusiness systemsModelsFiles + messages

One inspectable run

Follow the request from current authority to a confirmed result.

Maya starts a renewal from Codex. The run record shows what the source returned, what AI received, the exact approved change, how Onterix reconciled execution, and the destination receipts.

REQ-02418CUSTOMER RENEWAL · RUNTIME RECORD

Confirmed record

Requester: Maya Chen · person
Started in: Codex · MCP client
Capability: Customer renewal · v3.2

01Request

“Prepare the Acme renewal brief, advance the opportunity from stage 2 to stage 3, and send the customer follow-up.”

Prepare a cited renewal brief
Advance the Salesforce opportunity from stage 2 to stage 3
Send the reviewed customer follow-up

ONTERIX RUNTIMESource access · data release · approval · execution · evidence

02Runtime controls

Current authority: Maya Chen · delegated Salesforce access · renewal policy v7
Data retrieved: Account, opportunity, buyer, renewal date, and private note · 5 fields
What AI receives: Account, stage, buyer, and renewal date · private note withheld · release rule v4
Exact approval: Opportunity stage 2 → 3 · reviewed email to the named buyer
Approval binding: Revenue Operations · old and new values, recipient, and request digest
Safe execution: Read back CRM and check email status before any retry

03Result confirmed

Brief: Ready · four source citations · private note absent
CRM: Stage 3 confirmed · Salesforce receipt retained
Email: Accepted once · provider receipt retained

The same capability can run for an autonomous agent or application under its own identity, permissions, and trigger. Each run gets its own inspectable record.

From job to capability

A capability is the reusable definition behind the job.

In Onterix, a capability keeps company context, a reviewed method, runtime controls, and the result together. Improve that definition once without rebuilding the job for every AI client, agent, application, or workflow.

CONTEXT

Context

Current company knowledge with source access, citations, freshness, and ownership intact.

Explore knowledge

METHOD

Method

The steps, tools, and tests the job follows, reviewed and owned by an accountable person.

Explore skills

CONTROLS

Controls

Rules for what may be retrieved, what each recipient receives, which changes may run, and when approval is required.

Explore data release

RESULT

Result

A defined outcome plus evidence of the data released, changes executed, and result confirmed.

Explore safe actions

Apply the architecture to a complete customer workflow, review the security model, or choose where the runtime operates.

See a customer workflow Review the security model Compare deployment modes

Connect the first system

Start with one recurring job and the systems it needs.

Connect a source, test the job as a private capability, and inspect the first result before you share it.

Connect your first system Plan an enterprise rollout

Define one trusted job. Run it from every AI interface.