The agentic platform for your business and engineering, and the foundation you build on

Intelligence you can govern.

Every agent, agent team and AI workflow runs under one identity, permission, approval and audit layer, and it is yours to build on.

An agent asks

Reconcile March supplier invoices

  • resolved: Identityacting for Ana, Finance team
  • passed: Permissionmay read supplier invoices
  • passed: Data classconfidential · within the ceiling
  • resolved: Modelyour on-prem model
  • passed: Approvalnot required for reading
  • passed: Auditrecorded, with its decision code
Allowed — every piece agreed
  • An agent never exceeds the person it acts for
  • Every decision recorded, with its reason
  • One permission model for people, agents and apps
  • Install only the packages your organisation needs
The agentic platform · agent teams

Agents that work as a team. Governed like everyone else.

Agent teams are part of your company’s agent layer, and people are in the team. The owner sets the goals, people review and decide, and agents do the work under the same rules. One agent is a demo; a team is a workforce: a lead that plans, specialists that own an area, reviewers that check work they did not write, and workers brought in for one task. How the team works is a harness strategy you install and version like code. No agent grades its own homework, and every step is in the audit log. Use the strategies we ship, shape them, or build your own.

A team is

  1. 01Persistent roles
  2. 02Ephemeral workers
  3. 03Harness strategies
  4. 04People alongside the agents
Build once

Build once. Reuse everywhere.

Your first process connects the systems it needs and sets its rules. Every process after it starts from there: the same sign-in and roles, the same approvals and audit, the same agents and connections. Each team builds only what is new.

Reuse never widens access. Each team, person and agent reaches a shared connection only within its own permissions.

One platform, many teams

What each new process starts with

Supplier management

Built with the first process

  • resolved: Sign-in, roles and permissions
  • resolved: Approvals and the audit trail
  • resolved: Agents that act within their owner’s authority
  • resolved: ERP and email connections
  • resolved: Supplier rules and screens
The first process sets up the shared pieces
Build on it

A new supplier-management application. Most of it is already there.

Identity, approvals, audit and agent execution come with the platform, and the connections your organisation has already made are reused. Your team builds the supplier rules and the screens. Your team builds the business; Xema is the foundation.

An authorised coding agent can build on the same foundation: it creates and extends applications inside the permissions you set. People and agents work on the same platform — delegating, coordinating and approving work under the same rules.

A new application

Supplier management

  • passed: Identityalready provided: sign-in, roles, permissions
  • passed: Approvalsalready provided, with the audit trail
  • passed: Agentsalready provided: agents run within their owner’s authority
  • passed: Connectionsreuse what is connected: ERP, mail, documents
  • resolved: Ruleswhat you build: supplier checks, payment terms, risk levels
  • resolved: Screenswhat you build: the views your buyers need
Your team builds the business part
Connect what you have

One request, many systems.

A customer asks to move a delivery date and correct the invoice. The CRM knows the order, the ERP knows the stock, Finance owns the invoice. Each system stays the system of record: your ERP stays your ERP.

Xema coordinates the work and the rules across them, including agents and automations from different vendors.

A customer asks

Move delivery to 14 March and correct the invoice

  1. passed: CRMfinds the customer and order SO-4471
  2. passed: ERPchecks stock and the new delivery date
  3. held: Financeapproves the invoice change; the work waits
  4. passed: Emailconfirms to the customer after approval

passed: Done — each system kept its own record

Govern

An agent wants to issue a €18,400 refund.

A support agent finds a duplicate charge and asks to refund it. Xema checks what the agent may do and what the person it acts for may do. Refunds above €10,000 need Finance, so the work waits for the controller, and runs once she approves.

The AI can start the work. It cannot go around your rules.

An agent asks

Refund invoice INV-2291 · €18,400

  1. done: AgentAsks to refund €18,400for a customer, acting for Ana in Finance
  2. done: XemaChecks authority and policyrefunds above €10,000 need Finance approval
  3. done: XemaHolds the workit waits for an answer; it does not restart
  4. done: PersonThe finance controller approvesfrom the approvals queue
  5. done: XemaRuns the refund oncein your billing system
Refunded — approval and decision reason recorded
Your choice of AI

Change the AI. Keep everything else.

A model, an MCP server or another vendor’s agent sits behind a capability you defined, so you can switch it without touching processes, permissions or connections.

Start with external services, move to your own models on your own servers, or run with no network at all. The architecture is the same at every stage.

Agents ask for a lane, never a model. Planner, coder, review, light: your organisation maps each lane to a model in one place. Swap a model for every agent with one config change, and bring your own provider keys.

You change the AI

Processes, permissions and connections stay

  • resolved: Laneagents ask for “coder”, never a model name
  • passed: Swapone config change moves every agent
  • passed: Providerswitched: processes, rules and connections unchanged
  • passed: On-premyour own model: the same rules, on your servers
  • passed: Air-gappedno network at all: the same rules again
Nothing to redesign
The agentic platform · engineering teams

Any agent, any runtime. One set of rules.

The agentic layer under all your agents: permissions, approvals and audit that stay the same whichever runtime and model do the work. The eight questions every team asks before agents reach production, answered in one place, with room to bring or build your own runtime.

What it covers

  1. 01Runtimes
  2. 02Orchestration
  3. 03Tool security
  4. 04Infrastructure
  5. 05Models
  6. 06Sandboxes
  7. 07Evals and audit
  8. 08Pipeline gates

Your teams work on the business. The rest is already done.

Real work, across your business and engineering

Six examples of what teams run on Xema.

Each is something a team builds and runs on Xema, connected to the systems the work already lives in, with a person answering where the decision matters.

Finance

Month-end close.

Bank, suppliers and ledger reconciled, anomalies sent to the controller, and nothing posted without sign-off.

Read more

Customer support

Cases resolved, not just answered.

A duplicate charge is confirmed, refunded in billing after approval, and the customer is told.

Read more

Contracts & legal

Redlines against your playbook.

Every clause checked against your standards; counsel approves only the deviations.

Read more

People & IT

Joiners and leavers.

Access granted by role on day one, sensitive systems approved by a manager, everything revoked on the last day.

Read more

Engineering

Software delivery with agents.

A ticket is planned, coded in a sandbox and reviewed against your secure-coding skill. The tech lead approves before release.

Read more

Operations

An operations desk that answers fast.

An agent recommends a maintenance window from live data. A supervisor approves the crew booking, the team hears in Teams or Slack, and every refusal is audited.

Read more
How we help you get live

A platform. And the team that gets it working.

Putting AI to work at the core of your business is part product, part programme. Xema is yours to run, and our team works alongside yours until the first process is live and your people own it.

Platform

Xema, the product

The foundation your solutions run on, yours to operate.

  • Cloud, on-prem or air-gapped
  • Connections to the systems you use
  • Grows by installing packages

Partnership

Our team, alongside yours

We scope the first process with you, build it with you, and hand it over.

  • Process and approval design
  • Connections and co-built packages
  • Handover to your own team
Get a demo

Start with one process. Build on it from there.

We work with a small number of teams each quarter. We take one real process, connect the systems it touches, and put it live under your own approvals. The next process reuses what the first one built, and anything you build on top, an app, an agent or a whole team, starts from the same foundation. One agentic platform for the whole company, and a foundation to build anything on, your own way.

How it goes

  1. 01Scope one real process
  2. 02Connect the systems it touches
  3. 03Put it live, with its approvals
  4. 04The next one reuses all of it

Same platform the whole way.

Get a demo
Xema — Intelligence you can govern.