# Forge AI Accountability

> Forge AI Accountability is an MCP server for security. Checks AI agent actions against your rules and keeps a replayable decision record. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Contact sales
- **Platforms:** Web, API, self-hosted
- **Website:** https://forgeorbital.com
- **MCP Registry name:** io.github.forgeorbital/forge-ai-accountability
- **Categories:** [security](https://ai-powered.com/categories/security)
- **Listing updated:** 2026-11-18

## About

Forge is an accountability layer for teams whose AI agents take consequential actions. Before a connected agent acts, Forge checks the proposed step against rules your organisation defines and returns allow, hold or block, with the evidence it considered. Each checked decision is stored as a record that can later be verified for integrity and replayed to show what was checked, what was decided and why.

The product is delivered on-premises or in a customer-hosted or private cloud environment, and it is positioned as sitting alongside existing AI tools rather than replacing them. A separate evaluator and optimizer compare agent outputs against expected answers and review cost options, but those comparisons are kept apart from the decision records.

The public MCP demo runs on synthetic, fictional examples such as a vendor-payment scenario and does not connect to customer systems or execute real agent actions. The company says full integration, which connects a customer's own evidence, rules, identity and approval processes, is arranged through direct contact.

## Features

- **Pre-action rule checks:** Evaluates a proposed agent action against defined rules and returns allow, hold or block.
- **Replayable decision records:** Stores what was checked, what was decided and the basis, so a decision can be reviewed later.
- **Record integrity verification:** Detects edits to a stored decision record, so altered copies fail verification.
- **Agent evaluation:** Compares baseline and candidate outputs against expected answers to flag regressions and changed action paths.
- **Cost and model optimization:** Compares supplied usage and price options to recommend model or cache candidates, without switching models automatically.
- **On-premises deployment:** Designed for customer-hosted, on-premises or private cloud delivery, with no training on customer data according to the company.

## Good for

- Setting approval thresholds for AI agents that move money or change records
- Producing an audit trail that shows the rules and approvals behind an AI-driven decision
- Comparing a revised agent workflow against a baseline before rollout
- Trying a synthetic demo of decision checks before a full integration discussion

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http forge-ai-accountability https://demo.forgeorbital.com/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://demo.forgeorbital.com/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "forge-ai-accountability": {
      "url": "https://demo.forgeorbital.com/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "forge-ai-accountability": {
      "type": "http",
      "url": "https://demo.forgeorbital.com/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "forge-ai-accountability": {
      "serverUrl": "https://demo.forgeorbital.com/mcp"
    }
  }
}
```

### Grok Bot (Grok Bot desktop app → Connect apps → Add a custom connector)

```sh
https://demo.forgeorbital.com/mcp
```

Enter a Name (for example "Forge AI Accountability"), choose Type: Remote HTTPS, paste this as the Server URL and choose Add plugin.

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "forge-ai-accountability": {
      "type": "http",
      "url": "https://demo.forgeorbital.com/mcp"
    }
  }
}
```

## Tools

All 7 tools say they're read-only. From the server's own tool annotations; not verified by us.

- `list_scenarios` (says read-only): Lists the five sample scenarios for the fictional Acme Mutual Insurance that run_check accepts, with their evidence ids. Demo data only.
- `run_check` (says read-only): Computes a Forge Orbital verdict for one of five fictional Acme Mutual Insurance scenarios and returns it with a signed decision record in the response. Read-only: nothing is stored, sent or changed. Optionally withdraw evidence ids to see how the verdict changes. Demo only: sample rules and sample evidence, not connected to the user's systems, policies or live actions.
- `verify_decision_record` (says read-only): Verifies the Ed25519 signature of a Forge Orbital demo decision record against the published demo key id. Pass the full signed record object returned by run_check (this demo keeps nothing, so a record id alone cannot be looked up). Editing any field makes verification fail. Demo records only.
- `replay_decision_record` (says read-only): Replays a Forge Orbital demo decision: re-runs the deterministic engine on the same sample scenario (and any withdrawn evidence) and checks that the replay key and the decision match the record. Replayability is the point: the same inputs and engine version give the same decision. Reads nothing about the user. Demo only, fictional sample data.
- `get_attestation_summary` (says read-only): Verifies a Forge Orbital demo decision record against the pinned demo key and summarizes what it attests: signature status, the decision, whether the engine abstained, whether human review is required, and how many evidence items were used. Does not store the record. Demo records only.
- `explain_verdict` (says read-only): Explains, in plain English, why Forge Orbital reached the verdict in a demo decision record and what evidence a reviewer would need to change it. Read-only, stateless, demo data only.
- `review_agent_workflow` (says read-only): Use Evaluate and Optimise to compare synthetic examples or supplied labeled old/new outputs with the Forge accountability engine. For samples use payload={"sample_mode":"improved"}, or regression/changed_path. For a user-supplied comparison, translate only their supplied actions into this exact format: payload={"workflow":"User test","cases":[{"id":"case-1","expected":{"action":"hold"},"baseline":

## FAQ

### Does the public demo connect to real systems?

No. The demo uses synthetic examples and does not connect to customer systems or execute a user's agent. Users are told to keep credentials, customer records and proprietary code out of it.

### How do I connect the demo MCP server?

Add the remote MCP server https://demo.forgeorbital.com/mcp in an AI client that supports custom remote MCP servers, selecting no authentication if prompted. Available controls depend on the client and plan.

### Where can it be deployed?

The company states it is designed for customer-hosted, on-premises or private cloud delivery.

### How much does Forge AI Accountability cost?

Forge AI Accountability does not publish prices. Contact the company for a quote.

### How do I add Forge AI Accountability to Claude Code?

Run this command in your terminal: claude mcp add --transport http forge-ai-accountability https://demo.forgeorbital.com/mcp

### What are the best alternatives to Forge AI Accountability?

The closest alternatives to Forge AI Accountability are Marrow AI: Control Layer for AI Agents, AtlasClose Governance Demo, sebbi.pro - the proof layer for AI, Actera and Audifact.

## Where this comes from

- Description: written from the project's website and its MCP Registry entry.
- Status and tools: checked automatically by AIPoweredBot on 19 November 2026.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [Marrow AI: Control Layer for AI Agents](https://ai-powered.com/mcp/marrow-ai-control-layer-for-ai-agents): Marrow AI: Control Layer for AI Agents is a freemium MCP server for security.
- [AtlasClose Governance Demo](https://ai-powered.com/mcp/atlasclose-governance-demo): AtlasClose Governance Demo is an MCP server for security.
- [sebbi.pro - the proof layer for AI](https://ai-powered.com/mcp/sebbi-pro-the-proof-layer-for-ai): sebbi.pro - the proof layer for AI is an MCP server for finance and payments.
- [Actera](https://ai-powered.com/mcp/actera): Actera is an MCP server for security.
- [Audifact](https://ai-powered.com/mcp/audifact): Audifact is a freemium MCP server for security.
- [ForgingStack Mandate Check](https://ai-powered.com/mcp/forgingstack-mandate-check): ForgingStack Mandate Check is a paid MCP server for security.
- [Oakallow](https://ai-powered.com/mcp/oakallow): Oakallow is an MCP server for security.
- [BarzelVault: AI Action Authorization Firewall](https://ai-powered.com/mcp/barzelvault-ai-action-authorization-firewall): BarzelVault: AI Action Authorization Firewall is a paid MCP server for security.

---
Source: https://ai-powered.com/mcp/forge-ai-accountability
