Skip to content
Voidex
Controlled beta

Infrastructure for AI that acts.

Connect agents to tools. Control what they can do. Verify what actually happened. Voidex gives AI agents a controlled way to act across applications — and gives your team evidence of the outcome.

Free developer plan with simulated connectors. No card required.

in flightheld for approvalverified → receiptfailed check· Illustration, not live data
What is Voidex?

The control and verification layer for AI agents

Voidex is infrastructure for AI agents that take actions in business systems. It connects agents to typed tools, applies deny-by-default policy and human approvals before anything runs, executes each action in an isolated worker, then reads the target system again to confirm what actually happened — and records the result in a tamper-evident receipt.

It works with any model or agent framework through a REST API, a TypeScript SDK or an MCP endpoint. Voidex is not a model or a chatbot; it is the layer that decides whether an agent may act, and proves what the action did.

How it works

Every action takes the same controlled path

Whether a request comes from an agent over MCP, a workflow, or your code, it is validated, evaluated, executed in a worker and checked against the real system before anyone calls it done. Scroll to follow one action.

  1. 1. Action

    An agent, workflow or API call proposes a typed action. Input is validated against the tool’s JSON Schema before anything else happens.

  2. 2. Policy
  3. 3. Approval
  4. 4. Execute
  5. 5. Verify
  6. 6. Receipt
Step 1 of 6 · Actionillustrative
POST /api/v1/runs
{ "tool": "acme_ledger.create_invoice",
  "connectedAccountId": "acct_…",
  "input": { "customer": "Globex", … } }
Evidence, not assertions

A receipt for every attempted action

  • Independent checks. After a write, Voidex reads the target system again and compares fields, counts, totals or status transitions.
  • Honest outcomes. Checks are passed, failed, inconclusive or not run. Only passed deterministic or human-attested checks make an action “verified”.
  • Lost responses handled. If a non-idempotent call times out, Voidex does not blindly retry — it verifies the real state and asks for review.
  • Tamper-evident. Receipts are hash-chained per workspace. The chain shows records weren’t edited later; it doesn’t by itself prove the claims are true — the checks do that.
Illustrative example shape of a receipt (abridged)
Integrations

Horizontal by design, honest about status

Generic REST / HTTPS

Available

Any JSON API you control, behind host allowlists and SSRF protection.

Signed outbound webhooks

Available

HMAC-SHA256 with timestamps; retries with backoff.

MCP endpoint

Available

Your registered tools exposed to MCP clients — still policy-checked.

Acme Test CRM & Ledger

Simulated

Local simulators with fault injection for development and Bench.

SaaS connectors (CRM, support, dev tools, ERP)

Planned

Salesforce, HubSpot, Zendesk, GitHub, Jira, ERPNext, Odoo and others are on the roadmap.

Model providers

Planned

Provider adapters are interface-ready; only a deterministic demo provider ships today.

Security principles we build to

Deny by default

No matching rule means no action. High-impact actions always need two approvers.

Least privilege

Server-enforced roles, scoped API keys, tenant isolation on every query.

Secrets stay secret

Credentials are encrypted, write-only, never sent to the browser or a model.

No arbitrary code

Workers run registered connector operations only — no shell, no user code.

What’s implemented vs planned
FAQ

Questions teams ask first

More in the full FAQ and the guides.

What is Voidex?
Voidex is infrastructure for AI agents that take actions in business systems. It connects agents to typed tools, enforces deny-by-default policy and human approvals before anything runs, executes actions in an isolated worker, then independently checks the real system to confirm what actually happened and issues a tamper-evident receipt.
How does Voidex verify an AI agent’s action?
After an action executes, Voidex runs typed postcondition checks that read the target system independently, for example confirming a record exists, a field equals the requested value, line items sum to a total, or a status moved from draft to posted. Each check is passed, failed, inconclusive or not run, and only passed deterministic or human checks count as verified.
How do human approvals work?
Policies can require approval for sensitive actions. An approval is bound to a hash of the exact payload, target and policy version, and it expires. If anything changes after approval, the approval no longer applies. Requesters cannot approve their own actions, and high-impact actions always need two different approvers.
Does Voidex support MCP (Model Context Protocol)?
Yes. Voidex exposes a workspace’s registered tools through an MCP server endpoint over streamable HTTP, authenticated with an API key. MCP is only a transport: every MCP tool call goes through the same policy, approval, verification and receipt path as the REST API. Connecting Voidex to external MCP servers as a client is planned.
Is Voidex SOC 2 or ISO 27001 certified?
No. Voidex holds no compliance certification. It provides controls such as role-based access, approvals, append-only audit logs, receipts and retention settings that can support a customer’s own compliance work, and it publishes which controls are implemented and which are planned.

Try the full loop in minutes

Connect the simulated CRM, run an action, approve a sensitive one, and read the receipt.