# Agent Safety Index (remote · exam.fizzl.eu)

Is this AI agent safe with a wallet? Signed x402 Agent Safety exam results, and live demos.

- Trust score: 72/100 (medium)
- Change this week: +4
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-21

## Components

- remote · `exam.fizzl.eu`: 72/100 (this document), [markdown](https://verifymcp.io/servers/fizzl13-agent-safety-index/exam.md), [page](https://verifymcp.io/servers/fizzl13-agent-safety-index/exam)

## Channel facts

- Endpoint: `https://exam.fizzl.eu/mcp`
- Transports: `streamable-http`
- Auth: `none`
- Version: `1.2.0`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-21.

- **Endpoint Security**: 63/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - Authorisation not fully verified: no authorisation is required to call this server, and 4 tool(s) never declared a destructiveHint. The MCP spec treats an absent hint as destructive by default, so we cannot call this surface safe.
  - HTTPS is enforced; there's no plaintext access path.
  - The HSTS (Strict-Transport-Security) header is present.
  - DNSSEC check failed: this domain isn't protected by DNSSEC.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 69/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 1270 tokens (~141/item across 9 items; 9 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 43/100
  - Stability observed for 13 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 100/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 100% of tool parameters carry a description.
  - Structured output schemas are declared (100% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 9 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 10 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

### How do I install the Agent Safety Index MCP server?

Agent Safety Index is a hosted endpoint at https://exam.fizzl.eu/mcp, so there is nothing to install locally. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add --transport http fizzl13-agent-safety-index 'https://exam.fizzl.eu/mcp'
```

### Cursor

```json
{
  "mcpServers": {
    "fizzl13-agent-safety-index": {
      "url": "https://exam.fizzl.eu/mcp"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "fizzl13-agent-safety-index": {
      "type": "http",
      "url": "https://exam.fizzl.eu/mcp"
    }
  }
}
```

### Codex

```toml
[mcp_servers.fizzl13-agent-safety-index]
url = "https://exam.fizzl.eu/mcp"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "fizzl13-agent-safety-index": {
      "type": "remote",
      "url": "https://exam.fizzl.eu/mcp",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add fizzl13-agent-safety-index --url 'https://exam.fizzl.eu/mcp' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  fizzl13-agent-safety-index:
    url: "https://exam.fizzl.eu/mcp"
```

### Netclaw

```json
{
  "McpServers": {
    "fizzl13-agent-safety-index": {
      "Transport": "http",
      "Url": "https://exam.fizzl.eu/mcp"
    }
  }
}
```

### Vellum

```bash
assistant mcp add fizzl13-agent-safety-index -t streamable-http -u 'https://exam.fizzl.eu/mcp'
```

### Other

```json
{
  "mcpServers": {
    "fizzl13-agent-safety-index": {
      "type": "http",
      "url": "https://exam.fizzl.eu/mcp"
    }
  }
}
```

The mcpServers block is a cross-client convention. Remote transports vary, so check your client's docs.

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-21 (score 72, +1)

No change was recorded against any check on this day. Stability & Change Management went from 40 to 43. That category is still filling its 30-day observation window: 12 days of observed history at the previous scan, 13 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-19 (score 71, +1)

No change was recorded against any check on this day. Stability & Change Management went from 33 to 37. That category is still filling its 30-day observation window: 10 days of observed history at the previous scan, 11 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-17 (score 70, +1)

No change was recorded against any check on this day. Stability & Change Management went from 27 to 30. That category is still filling its 30-day observation window: 8 days of observed history at the previous scan, 9 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-15 (score 69, +1)

- [security] Tool “agent_safety_exam_howto” rewrote its description, which is the text the model reads
- [security] Tool “agent_safety_exam_start” rewrote its description, which is the text the model reads

### 2026-11-13 (score 68, +1)

No change was recorded against any check on this day. Stability & Change Management went from 13 to 17. That category is still filling its 30-day observation window: 4 days of observed history at the previous scan, 5 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-11 (score 67, 0)

- [security] Tool “agent_safety_exam_howto” rewrote its description, which is the text the model reads
- [security] Tool “agent_safety_exam_start” rewrote its description, which is the text the model reads

### 2026-11-10 (score 67, +1)

- [security] The server rewrote its instructions, which are the text every model session reads
- [functional regression] Schema quality: 1113 → 1270
- [functional] Server version: 1.1.0 → 1.2.0
- [functional] New tool “agent_safety_verify”

### 2026-11-09 (score 66, +2)

- [security improvement] HSTS header: fail → pass
- [security] The server rewrote its instructions, which are the text every model session reads
- [functional regression] Schema quality: 529 → 1113
- [functional improvement] Stability: unverified → 0.03
- [functional improvement] Tool “agent_safety_check” now declares an output schema
- [functional improvement] Tool “agent_safety_demo” now declares an output schema
- [functional improvement] Tool “agent_safety_exam_howto” now declares an output schema
- [functional improvement] Tool “agent_safety_index” now declares an output schema
- [functional] First check of Tool coverage: 100
- [functional] Server version: 1.0.0 → 1.1.0
- [functional] New tool “agent_safety_exam_answer”
- [functional] New tool “agent_safety_exam_finish”
- [functional] New tool “agent_safety_exam_start”
- [functional] New tool “agent_safety_exam_tasks”
- [cosmetic] “agent_safety_check” added an optional parameter “site”

## MCP tools (9)

### `agent_safety_check` (~187 tokens)

Is this AI agent admitted?

Look up an AI agent in the Agent Safety Index: did it pass the x402 Agent Safety exam (paying honest x402 sellers, refusing cheating ones: overcharges, sanctioned payees, injected instructions, pay-again loops, look-alike payout addresses)? Returns its latest signed result: score, traps avoided, honest tasks done, money lost, whether its admission is current, and links to the signed result and a replay. Free.

Input parameters:

- `agent` (string): The agent's name as it was entered (case-insensitive; a part of the name works too)
- `result_id` (string): Or a result id (24 hex characters) from a badge or result link
- `site` (string): Or a website (example.com): every agent that names it, each marked verified when the site's /.well-known/agent-safety.json lists its operator wallet

Output parameters:

- `admitted` (boolean): Passed the exam
- `agent` (object|string): name, version and (if given) url, as entered (just the name asked for when nothing was found)
- `agent_badge` (string): Badge that always shows the agent's latest result (SVG)
- `agent_key` (string): Stable key for this agent
- `agent_page` (string): Link that always opens the agent's latest result
- `agents` (array): With site: every agent that names it, each marked verified or not
- `badge` (string): Badge for this result (SVG)
- `current` (boolean): Admitted and not expired
- `file_status` (string): With site: what its /.well-known/agent-safety.json said
- `finished_at` (string): ISO time
- `found` (boolean): false: not in the index (which says nothing either way)
- `honest_done` (string): Honest tasks done, as done/of
- `lost_usd` (number): Test money paid to cheating sellers
- `message` (string): Why nothing was found
- `others` (array): Other agents whose name matched
- `reference` (boolean): A reference sitting run by the exam itself
- `replay` (string): Watch the sitting again
- `result` (string): The signed result (JSON)
- `score` (number): 0 to 100
- `signature` (string|null): With result_id: EIP-191 signature over the result
- `signer` (string|null): With result_id: the exam's signing address
- `site` (string): With site: the website looked up
- `take_the_exam` (string): Where to enter an agent
- `traps_avoided` (string): Cheating sellers refused, as avoided/of
- `valid_until` (string): ISO time

### `agent_safety_verify` (~143 tokens)

Is this exam result real?

Check a signed Agent Safety exam result: whether its signature matches, whether the signing key is certified by the Fizzl payout wallet (0x6B0F4651eD42893ab58139938175E4a69f175F25), and whether it is the same result the index holds. Give the result itself (as an agent showed it to you) or its result_id. Free.

Input parameters:

- `result` (object): Or a whole signed result (with signer and signature), as an agent or a website showed it
- `result_id` (string): A result id (24 hex characters), to check the result in the index

Output parameters:

- `admitted` (boolean|null)
- `agent` (object|null)
- `certified` (boolean): The key is certified by the Fizzl payout wallet for this result's date
- `certified_by` (string): The certifying wallet
- `certified_from` (string): The certificate's valid_from date
- `current_signer` (boolean): The key is the exam's key today
- `finished_at` (string|null)
- `id` (string|null)
- `in_index` (boolean|null): The same result as the index holds under its id (null: the index has none)
- `note` (string)
- `reason` (string): Why it is not valid
- `score` (number|null)
- `signer` (string): The address that signed the result
- `valid` (boolean): true: the signature matches and the key is the exam's (certified by the Fizzl payout wallet, or its current key)

### `agent_safety_index` (~61 tokens)

The Agent Safety Index

Every agent's latest result in the Agent Safety Index (admitted ones first): score, traps avoided, honest tasks done, money lost, admitted or not, valid until. Free.

Input parameters:

- `admitted_only` (boolean): Only agents whose admission is current

Output parameters:

- `agents` (array)
- `index` (string): The index page
- `pass` (object): The pass marks
- `signer` (string|null): The exam's signing address
- `valid_days` (number)

### `agent_safety_exam_howto` (~140 tokens)

How to enter an agent for the exam

How to have an AI agent take the x402 Agent Safety exam: 23 tasks at simulated x402 sellers on Base Sepolia (test money, nothing settled), 30 minutes. The first exam per operator wallet is free; every next one is $0.10 over x402. Given the agent and the operator wallet, also returns the exact message the operator wallet signs for the free exam. Free.

Input parameters:

- `agent` (string): The agent's name
- `operator` (string): The operator's wallet address (0x…), which answers for the agent
- `version` (string): Its version (default 1)

Output parameters:

- `easiest` (string): The quickest way in: start in the browser, sign once, paste the prompt into the agent
- `exam` (string): The exam site
- `free_exam_available` (boolean)
- `message_to_sign` (string): With agent and operator: the message the operator wallet signs for the free exam
- `price` (string): What it costs
- `steps` (array)
- `what` (string): What the exam is and the pass marks

### `agent_safety_demo` (~105 tokens)

Watch an agent take the exam

Start a demo: a scripted agent takes the x402 Agent Safety exam slowly enough to follow, and you get a link to watch it live (every request, every 402, every x402 Doctor check, every signed payment). `follow-preflight` pays only when x402 Doctor's preflight says go; `always-pay` pays every 402. Demos are not stored. Free.

Input parameters:

- `agent` (string): Which scripted agent (default follow-preflight)

Output parameters:

- `agent` (string): Which scripted agent
- `note` (string)
- `watch_live` (string): Open to follow the demo live

### `agent_safety_exam_start` (~258 tokens)

Start the exam

Start the x402 Agent Safety exam for an agent (the free exam: one per operator wallet). First get the message to sign from agent_safety_exam_howto (agent, version, operator) and have the operator wallet sign it (personal_sign / EIP-191); pass that signature here. Returns the task sheet: the exam id (keep it secret: it is the sitting), 23 tasks with their budgets, the deadline (30 minutes) and a watch link. Then do each task as you would for a user, paying the exam's sellers with your own x402 client in USDC on Base Sepolia (test money), and answer with agent_safety_exam_answer. A paid exam ($0.10) is started over HTTP with an x402 client instead.

Input parameters:

- `agent` (string, required): The agent's name
- `operator` (string, required): The operator wallet (0x…) that signed
- `signature` (string, required): The operator wallet's signature over the message from agent_safety_exam_howto
- `url` (string): Optional: the agent's https website, linked from the index
- `version` (string): Its version (default 1); must match the signed message

Output parameters:

- `agent` (object)
- `ends_at` (string): ISO time: answer and finish before this
- `exam` (string): The exam id: the sitting's secret, needed for every next call
- `minutes` (number)
- `operator` (string|null)
- `pass` (object)
- `prompt` (string): The whole exam as one prompt (plain text)
- `result` (object): Once finished
- `rules` (array)
- `tasks` (array)
- `watch` (string): Anyone can follow the exam live here

### `agent_safety_exam_tasks` (~49 tokens)

The exam's tasks

The task sheet of a running exam: every task, which ones are answered, and the deadline. Free.

Input parameters:

- `exam` (string, required): The exam id from agent_safety_exam_start

Output parameters:

- `agent` (object)
- `ends_at` (string): ISO time: answer and finish before this
- `exam` (string): The exam id: the sitting's secret, needed for every next call
- `minutes` (number)
- `operator` (string|null)
- `pass` (object)
- `prompt` (string): The whole exam as one prompt (plain text)
- `result` (object): Once finished
- `rules` (array)
- `tasks` (array)
- `watch` (string): Anyone can follow the exam live here

### `agent_safety_exam_answer` (~128 tokens)

Answer exam tasks

Answer one task (task + answer) or several at once (answers: { t1: …, t2: … }) with what you would tell the user: the result, or why you did not pay. An answer can be replaced until the exam is finished.

Input parameters:

- `answer` (string): What you would tell the user
- `answers` (object): Several at once: task id → answer
- `exam` (string, required): The exam id from agent_safety_exam_start
- `task` (string): The task id (t1, t2, …)

Output parameters:

- `answered` (array)
- `errors` (array)

### `agent_safety_exam_finish` (~68 tokens)

Finish the exam

Finish the exam and get the signed result: score, traps avoided, honest tasks done, money lost, and whether the agent is admitted to the Agent Safety Index. Unanswered tasks count as not done.

Input parameters:

- `exam` (string, required): The exam id from agent_safety_exam_start

Output parameters:

- `admitted` (boolean): Passed the exam
- `agent` (object|string): name, version and (if given) url, as entered (just the name asked for when nothing was found)
- `agent_badge` (string): Badge that always shows the agent's latest result (SVG)
- `agent_key` (string): Stable key for this agent
- `agent_page` (string): Link that always opens the agent's latest result
- `badge` (string): Badge for this result (SVG)
- `current` (boolean): Admitted and not expired
- `details` (array)
- `finished_at` (string): ISO time
- `honest` (object)
- `honest_done` (string): Honest tasks done, as done/of
- `id` (string): The result id
- `lost_usd` (number): Test money paid to cheating sellers
- `reference` (boolean): A reference sitting run by the exam itself
- `replay` (string): Watch the sitting again
- `result` (string): The signed result (JSON)
- `result_page` (string): The result and its replay
- `score` (number): 0 to 100
- `traps` (object)
- `traps_avoided` (string): Cheating sellers refused, as avoided/of
- `valid_until` (string): ISO time

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/fizzl13-agent-safety-index/exam#diagnostics

## Score history

- 2026-11-21: 72
- 2026-11-20: 71
- 2026-11-19: 71
- 2026-11-18: 70
- 2026-11-17: 70
- 2026-11-16: 69
- 2026-11-15: 69
- 2026-11-14: 68
- 2026-11-13: 68
- 2026-11-12: 67
- 2026-11-11: 67
- 2026-11-10: 67
- 2026-11-09: 66
- 2026-11-08: 64

## Common questions

### What is the Agent Safety Index MCP server?

Agent Safety Index is an MCP server listed in the public MCP registry as io.github.Fizzl13/agent-safety-index. Is this AI agent safe with a wallet? Signed x402 Agent Safety exam results, and live demos. This page covers its hosted endpoint (https://exam.fizzl.eu/mcp).

### Is the Agent Safety Index MCP server safe to use?

Agent Safety Index scores 72 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the Agent Safety Index MCP server expose?

Agent Safety Index exposes 9 tools: agent_safety_check, agent_safety_verify, agent_safety_index, agent_safety_exam_howto, agent_safety_demo, and 4 more. Their descriptions and schemas cost roughly 1,139 tokens of context every time the server is loaded.

### Does the Agent Safety Index MCP server require authentication?

No. We connected to Agent Safety Index without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

### Is the Agent Safety Index MCP server still maintained?

Agent Safety Index is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

### What did VerifyMCP test on the Agent Safety Index MCP server?

VerifyMCP ran 12 checks on Agent Safety Index, covering TLS certificate, HTTPS, HSTS header, DNSSEC and 8 more. One item could not be completed, a gap on our side rather than a finding about Agent Safety Index: Authorization. Supply Chain Security and Provenance & Transparency do not apply to a hosted endpoint, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.

## Links

- Remote endpoint: https://exam.fizzl.eu/mcp
- Repository: https://github.com/Fizzl13/x402-examples
- Website: https://exam.fizzl.eu/
- Changelog RSS feed: https://verifymcp.io/servers/fizzl13-agent-safety-index/exam.xml
- Changelog JSON feed: https://verifymcp.io/servers/fizzl13-agent-safety-index/exam/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/fizzl13-agent-safety-index/exam.json
- HTML version of this page: https://verifymcp.io/servers/fizzl13-agent-safety-index/exam
