Build AI-native teams.
How do you want to scale with AI?
Startup — Move faster than incumbents.
SMB — Automate without losing control.
Enterprise — Lead the AI shift in your department.
For startups
Compete with teams ten times your size.
- Give every team member their own team of AI agents.
- Agents handle everything that isn't your edge.
- Fully controlled by the team member who runs them.
Traditional governance is built for employees.
- Role hierarchies. Service accounts. IT-managed permissions.
- Designed for people who log in and click through tools.
The current approach: treating agents like employees.
- Agents get service accounts and credentials. IT manages them like employees.
- Rigid workflows define what each agent can do — step by step.
- It doesn't scale. Nobody's accountable.
With HAP, agents extend each team member.
- Not "what can this agent do" — but "what do you want to achieve, and how much autonomy do you trust?"
- Agents act as extensions of the team member who authorized them.
- Team members authorise. HAP issues a receipt for every action.
Governed by the HAP authorisation server.
Each team member defines the intent and bounds for autonomous execution. The authorisation server enforces it.
- Low risk — Agents execute autonomously — no approval needed.
- High risk — Agents pause for approval.
- Critical — Multiple team members must approve.
The authorisation server issues a receipt for every AI agent action.
No receipt. No execution.
A team member approves what an agent is allowed to do. The system enforces it. Every action produces a receipt anyone can check.
- Authorisation Server — where a team member records their approval, along with the exact limits.
- HAP Gateway — checks the approval before any action runs. Blocks anything outside the approved limits.
- Agent — runs the action, but only if the Gatekeeper allows it.
Open protocol. No lock-in.
- Based on the open-source Human Agency Protocol.
- The authorisation server signs — it never sees the semantics.
- Extend with open-source MCP connectors.
- Your data. Your audit trail. Yours.
Extend with MCP — the AI agent interface standard.
- Most services have their own MCP server.
- Growing community of MCP connectors.
- The local HAP Gateway adds more and more out of the box.
Out of the box: Stripe, Mollie, Gmail, LinkedIn, CRM, Records.
How to get started
- Use your own AI agents — Claude Code, Codex, Cursor, Openclaw, or any MCP-compatible agent.
- Create Account here — Set up your team on the HAP authorisation server. Authorise agents. Issue receipts.
- Download the local Gateway — Open-source gateway that runs on your machine. Sits between your agents and the tools they use.
- Extend — Connect your services and resources via MCP — the open standard for AI agent interfaces.
For SMBs
Growth is blocked by overhead.
- Every new hire adds ops, IT, admin load.
- Scaling the team means scaling the back office.
- The bigger you get, the slower you move.
Hire and drown, or automate and risk it.
- Hire more → overhead compounds, margin shrinks.
- Automate loosely → one wrong action reaches a customer.
- You need both growth and control.
Traditional governance is built for employees.
- Role hierarchies. Service accounts. IT-managed permissions.
- Designed for people who log in and click through tools.
The current approach: treating agents like employees.
- Agents get service accounts and credentials. IT manages them like employees.
- Rigid workflows define what each agent can do — step by step.
- It doesn't scale. Nobody's accountable.
With HAP, agents extend each team member.
- Not "what can this agent do" — but "what do you want to achieve, and how much autonomy do you trust?"
- Agents act as extensions of the team member who authorized them.
- Team members authorise. HAP issues a receipt for every action.
Governed by the HAP authorisation server.
Each team member defines the intent and bounds for autonomous execution. The authorisation server enforces it.
- Low risk — Agents execute autonomously — no approval needed.
- High risk — Agents pause for approval.
- Critical — Multiple team members must approve.
The authorisation server issues a receipt for every AI agent action.
No receipt. No execution.
A team member approves what an agent is allowed to do. The system enforces it. Every action produces a receipt anyone can check.
- Authorisation Server — where a team member records their approval, along with the exact limits.
- HAP Gateway — checks the approval before any action runs. Blocks anything outside the approved limits.
- Agent — runs the action, but only if the Gatekeeper allows it.
Open protocol. No lock-in.
- Based on the open-source Human Agency Protocol.
- The authorisation server signs — it never sees the semantics.
- Extend with open-source MCP connectors.
- Your data. Your audit trail. Yours.
Extend with MCP — the AI agent interface standard.
- Most services have their own MCP server.
- Growing community of MCP connectors.
- The local HAP Gateway adds more and more out of the box.
Out of the box: Stripe, Mollie, Gmail, LinkedIn, CRM, Records.
How to get started
- Use your own AI agents — Claude Code, Codex, Cursor, Openclaw, or any MCP-compatible agent.
- Create Account here — Set up your team on the HAP authorisation server. Authorise agents. Issue receipts.
- Download the local Gateway — Open-source gateway that runs on your machine. Sits between your agents and the tools they use.
- Extend — Connect your services and resources via MCP — the open standard for AI agent interfaces.
For enterprises
Traditional governance is built for employees.
- Role hierarchies. Service accounts. IT-managed permissions.
- Designed for people who log in and click through tools.
The current approach: treating agents like employees.
- Agents get service accounts and credentials. IT manages them like employees.
- Rigid workflows define what each agent can do — step by step.
- It doesn't scale. Nobody's accountable.
With HAP, agents extend each team member.
- Not "what can this agent do" — but "what do you want to achieve, and how much autonomy do you trust?"
- Agents act as extensions of the team member who authorized them.
- Team members authorise. HAP issues a receipt for every action.
Governed by the HAP authorisation server.
Each team member defines the intent and bounds for autonomous execution. The authorisation server enforces it.
- Low risk — Agents execute autonomously — no approval needed.
- High risk — Agents pause for approval.
- Critical — Multiple team members must approve.
The authorisation server issues a receipt for every AI agent action.
No receipt. No execution.
A team member approves what an agent is allowed to do. The system enforces it. Every action produces a receipt anyone can check.
- Authorisation Server — where a team member records their approval, along with the exact limits.
- HAP Gateway — checks the approval before any action runs. Blocks anything outside the approved limits.
- Agent — runs the action, but only if the Gatekeeper allows it.
Open protocol. No lock-in.
- Based on the open-source Human Agency Protocol.
- The authorisation server signs — it never sees the semantics.
- Extend with open-source MCP connectors.
- Your data. Your audit trail. Yours.
Your authorisation infrastructure. Your choice.
- Use HAP's hosted authorisation service — ready to go.
- Run the authorisation server in-house — full control over your data.
- Build your own — the protocol is open.
Extend with MCP — the AI agent interface standard.
- Most services have their own MCP server.
- Growing community of MCP connectors.
- The local HAP Gateway adds more and more out of the box.
Out of the box: Stripe, Mollie, Gmail, LinkedIn, CRM, Records.
Compliance, by design.
- EU AI Act Article 14 — human oversight, structural.
- ISO 42001 — every action has a decision owner.
- NIST AI RMF — tamper-proof trail.
Receipts as evidence.
- Who approved. What limits. What happened.
- Cryptographically signed.
- Verifiable without our servers.
Fits your identity stack.
- SAML, OIDC, your existing IdP.
- Agents don't get identities — nothing new in your IAM.
- Team members sign in with what they already use.
How to get started
- Use your own AI agents — Claude Code, Codex, Cursor, Openclaw, or any MCP-compatible agent.
- Create Account here — Set up your team on the HAP authorisation server. Authorise agents. Issue receipts.
- Download the local Gateway — Open-source gateway that runs on your machine. Sits between your agents and the tools they use.
- Extend — Connect your services and resources via MCP — the open standard for AI agent interfaces.