# Can an AI agent use HumanMirror? L3 · Does the work

HumanMirror sells safety checks and small data tools for AI agents.

**L3 · Does the work** · [What the levels mean](https://agentdrives.com/method/levels?via=md)

Rank among the 246 products we drove:

- Time to get in: 3 minutes, tied #29 (Getting in on its own)
- Holds up: 75 / 100, tied #65 (Reliability under errors)
- Core work done: 75%, tied #74 (6 of 8 core tasks)
- Steps on its own: 2 / 5, tied #86 (Find to Leave, no person)

**Can an AI agent use HumanMirror? Yes.** Our agent signed up for HumanMirror on its own over its API, got an API key and used its API and MCP server. No CLI found. It offers agents a way to pay: x402.

- Level confirmed: Not yet (Level from one drive; a second must agree)
- Got in by: Key by API (3 minutes)
- Got as far as: Leave: can't delete its account (Can pay on its own)

Category: Auth and identity · Connects by: MCP, API · Agent payments: x402; our agent used it · Driven: 15 Nov 2026

**How to connect**
- MCP server URL: `https://humanmirror.fr/api/x402/mcp/` (streamable HTTP, tested in our drive). Source: https://humanmirror.fr/api/x402/mcp/
- API base URL: `https://humanmirror.fr` (documented). Source: https://humanmirror.fr/x402/openapi.json

Checked 17 Nov 2026: MCP server answers, 15 tools; x402 payment challenge valid; sign-up page, llms.txt and OpenAPI unchanged since the check before.

See [how it ranks among auth providers for agents](https://agentdrives.com/research/best-auth-providers-for-ai-agents?via=md).

## From finding it to leaving

What an agent managed without a person, at each step.

1. **Find**: Partly. Docs with gaps: no getting-started page; has llms.txt, an OpenAPI and an MCP server. The level doesn't rest on Find: it counts getting in, using, paying and leaving. (12 doc pages read)
2. **Get in**: On its own. Key by API. Pay per call: stopped by a card. (3 minutes)
3. **Use**: Partly. Did: find the right HumanMirror service for a need, for free; remove duplicate records from a list of data. 2 not done, among them: repair broken JSON. (6 of 8 core tasks)
4. **Pay**: On its own. Valid x402 payment challenge: $0.001 a call. (1 way for agents to pay)
5. **Leave**: Partly. Nothing to cancel. Its account stays. No such route was found; erasure is requested by mail. First stop.

To reach L4: Do 80% of the core work (now 75%) and delete its account, so its keys and login stop working.

## How an agent connects

MCP, API, OpenAPI, agent payments and llms.txt found.

- **MCP server**: 6 of 16 tested. Official server. Two servers on 28 Jul 2026 (x402 catalog 1.5.0 and One 1.0.2), no auth for discovery. https://humanmirror.fr/api/x402/mcp/ (16 tools mapped in our drive) 1 tool listed without signing in on 3 Nov 2026: `humanmirror_flow`.
- **API**: 89 of 177 tested. REST JSON, Bearer trial keys and x402 v2 per-call payments. (177 operations mapped in our drive)
- **OpenAPI**: Yes. A machine-readable description of the API.
- **CLI**: None found.
- **SDK**: Unknown.
- **llms.txt**: Yes. https://humanmirror.fr/llms.txt
- **Agent payments**: Yes. x402 (Base). Plan and price: Nexus 5000-credit pack, €9.99. It offers x402 payment.
- **Docs**: Some gaps. Covers auth. Code examples on only some pages. No getting started or error codes. Describes each operation only in part: completeness 53 of 100. 4 of 17 documented ways to do the core work were tried, and 2 of them worked. (12 pages read)

[The standards](https://agentdrives.com/standard)

## Getting in

Our agent tried 2 of 4 ways in; 1 worked without a person.

The way that worked: **Key by API**, 3 minutes. No inbox, browser or person needed.

- Agent on a server (API, MCP, CLI): Partly. Signs up, gets an API key, makes its first API call and does 6 of its 8 core tasks, then stops: the rest needs a paid plan. It can't delete its account.
- Assistant in its own browser (Like ChatGPT agent): Partly. Signs up and does 6 of its 8 core tasks, then stops: the rest needs a paid plan. It can't delete its account.
- Signed agent (Known to Cloudflare): Yes. The site opens.

| Way in | What happened |
|---|---|
| Key by API | Worked |
| No key needed | Not reached |
| Pay per call | Stopped by a card |
| Try first, claim later | Not tried |

## What an agent can do with HumanMirror

6 of 8 core tasks done, and 18 of 24 others.

- Holds up: 75 of 100 (Reliability 100, Handles mistakes 50). The mean of the two.
- Docs match: 55 of 100 (Docs accuracy 56, Docs completeness 53). Docs accuracy counts 70%, completeness 30%.
- Surface tested: 80 of 100. 89 of 111 passed: the 202 documented operations less the 91 that spend money or reach real people

Core tasks:

| Task | Result | Done through | Tried, not done | Offered, not tried |
|---|---|---|---|---|
| Run a Safe Preflight check on an agent action | Done | API | – | MCP |
| Find the smallest matching capability for a need | Done | API, MCP | – | – |
| Issue an execution receipt after an agent action | Done | API | – | MCP |
| Execute a verified deterministic Outcome | Done | API | – | MCP |
| Deduplicate JSON records | Done | API | – | MCP |
| Run a 2-3 step verified Flow mission | Done | API | – | MCP |
| Clean and normalise an invalid JSON agent response | Partly | – | API | MCP |
| Run an Enterprise Guard preflight with a spend policy | Paid plan, not counted: They need a paid plan or a EUR 299 purchase. | – | API | – |
| Order an Agent Safety Audit of one workflow | Paid plan, not counted: They need a paid plan or a EUR 299 purchase. | – | – | Web app |
| Get an ALLOW / REVIEW / BLOCK verdict for one action | Blocked | – | MCP | API |

18 of 24 other tasks done:

- Scan a payload for leaked secrets: Done
- Normalise entity identifiers: Done
- Detect anomalies in a numeric series: Done
- Score data quality against a threshold: Done
- Quality-gate an AI output before it ships: Done
- Detect breaking JSON contract changes: Done
- Infer a JSON contract from a payload: Done
- Issue a Trace verification receipt: Done
- Verify a Trace receipt against a payload: Done
- Preflight a destructive operation with Circuit: Done
- Search the Nexus tool catalog: Done
- Route a tool call through Nexus: Done
- Sanitise untrusted agent context: Done
- Get free payment-rescue advice for a lost paid response: Done
- Qualify a mission in the OmniDome Mission Desk: Done
- Discover OmniDome Commons capabilities: Done
- Read the credit balance: Done
- Read live Execution Market state: Done
- Analyse a dataset with Oracle: Paid plan
- Publish a best-execution request on the Execution Market: Blocked
- Submit a provider bid on the Execution Market: Blocked
- Persist or recall shared inter-agent context: Blocked
- Check x402 payment compatibility before paying: Blocked
- Create a referral link: Blocked
- Buy a credit pack: Not tried

The plans they need:

- Create a referral link

The drive used the no-card trial.

What it can't do, and why:

- Repair broken JSON: The tool returned the broken text unchanged and still charged a credit.
- Run Enterprise Guard checks or order the safety audit: They need a paid plan or a EUR 299 purchase.
- Revoke a key or delete the account: No such route was found; erasure is requested by mail.
- Test the pay-per-call routes: Our own payment signer had reached its daily limit.

**Surface audit:** 202 documented operations: 99 run, 12 blocked by sandbox or plan limits and 91 left out because they spend money or reach real people. Of the 99 run, 89 passed and 10 didn't; 26 behaved differently from the docs. By surface, 89 on the API, 6 on the MCP server and 4 elsewhere, such as sign-in and webhooks.

## Tips for agent builders

- Start with POST /api/solve/ or the x402 MCP match tool to pick a service id, then run it through POST /api/nexus/call/ with the trial key instead of paying per x402 call.
- Paths in the OpenAPI specs lack trailing slashes; the live API 308-redirects, so call the slashed path.
- Many request bodies are undocumented: read the error code (objective_required, challenge_id_required) and add the named field.
- A 200 can still carry ok:false; check the ok field, not only the status.
- Failed Outcomes are refunded (charged_credits 0); retrying is cheap.

## Paying

Ways to pay checked 15 Nov 2026.

- Result: On its own. Valid x402 payment challenge: $0.001 a call.
- Checkout: Plan and price: Nexus 5000-credit pack, €9.99. It offers x402 payment.
- Ways to pay:
  - x402 (on its own domain · Base)

## Leaving

- Result: Partly. Nothing to cancel. Its account stays. No such route was found; erasure is requested by mail.
- Its account: Not on its own: blocked. No delete or revoke route was found in the 8 OpenAPI specs, the MCP tools or llms.txt; DELETE on the Nexus and OmniDome endpoints answered 405. The privacy page sends erasure requests by mail to privacy@humanmirror.fr, which our agent may not write to. Still there: the Nexus trial key, the Forge trial key, the OmniDome agent identity and key, and the OmniDome child agent.

## Checked every day

What our agent finds on the product's own site each day, without an account.

- Sign-up page, loads on the newest check; 3 of 3 days
- llms.txt, found on the newest check; 3 of 3 days
- MCP server, answers on the newest check (15 tools); 3 of 3 days
- OpenAPI, found on the newest check; 3 of 3 days
- Payment challenge, valid on the newest check (x402); 3 of 3 days

## How HumanMirror compares

In Auth and identity: 2 of the 9 products we drove in Auth and identity are higher; 5 others share L3. As of 15 Nov 2026.

- [FusionAuth](https://agentdrives.com/tools/fusionauth.md?via=md): L4, Gets in on its own
- [PropelAuth](https://agentdrives.com/tools/propelauth.md?via=md): L1, Can't sign up on its own
- [SuperTokens](https://agentdrives.com/tools/supertokens.md?via=md): L4, Gets in on its own
- [Clerk](https://agentdrives.com/tools/clerk.md?via=md): L3, Gets in on its own
- [WorkOS](https://agentdrives.com/tools/workos.md?via=md): L3, Gets in on its own
- [Auth0](https://agentdrives.com/tools/auth0.md?via=md): L3, Gets in on its own
- [Kinde](https://agentdrives.com/tools/kinde.md?via=md): L3, Gets in on its own
- **HumanMirror** (this page): L3, Gets in on its own
- [Web3Auth (MetaMask Embedded Wallets)](https://agentdrives.com/tools/web3auth.md?via=md): L3, Gets in on its own

[All 9 in Auth and identity](https://agentdrives.com/categories/auth?via=md)

[How we rank](https://agentdrives.com/method?via=md#order)

On our standard (33 criteria). Best or ahead on 11 of 33 criteria, behind on 9.

- [Pages fit one fetch](https://agentdrives.com/standard#docs-08): **The best result** All 7 others (auth and identity products) lower.
- [An agent payment rail on the core](https://agentdrives.com/standard#pay-02): **The best result** All 8 others (auth and identity products) lower.
- [A plan an agent can buy by API](https://agentdrives.com/standard#pay-03): **The best result** All 6 others (auth and identity products) lower.
- [Structured results](https://agentdrives.com/standard#mcp-08): **Ahead of 4 of 5** others (auth and identity products), 1 the same.
- [A request id on every answer](https://agentdrives.com/standard#err-07): **Ahead of 5 of 7** others (auth and identity products), 2 the same.
- [A path written for agents](https://agentdrives.com/standard#docs-10): **Ahead of 4 of 7** others (auth and identity products), 3 the same.
- [A person can claim what the agent made](https://agentdrives.com/standard#in-09): **Ahead of 13 of 23** others (every product), 10 the same.
- [A tool list that fits](https://agentdrives.com/standard#mcp-11): **Ahead of 48 of 91** others (every product), 43 the same.
- [Answers sized for an agent](https://agentdrives.com/standard#api-04): **Tied for the best result** 2 of 4 others (auth and identity products) the same, 2 lower.
- [Errors a program can read](https://agentdrives.com/standard#err-02): **Tied for the best result** 3 of 6 others (auth and identity products) the same, 3 lower.
- [The best way in is built for agents](https://agentdrives.com/standard#in-12): **Level with 4 of 7** others (auth and identity products), 3 lower.
- [Results sized for a context window](https://agentdrives.com/standard#mcp-09): **Ahead of 3 of 5** others (auth and identity products), 1 the same, 1 higher.
- [Retried payments charge once, with a receipt](https://agentdrives.com/standard#pay-04): **Level with 18 of 30** others (every product), 8 lower, 4 higher.
- [A way in with no person](https://agentdrives.com/standard#in-01): **Level with 7 of 8** others (auth and identity products), 1 lower.
- [Tool names that sort themselves](https://agentdrives.com/standard#mcp-03): **Level with 85 of 91** others (every product), 6 lower.
- [llms.txt that works](https://agentdrives.com/standard#find-02): **Level with 6 of 8** others (auth and identity products), 1 lower, 1 higher.
- [Fast to the first call](https://agentdrives.com/standard#in-03): **The same as all** 6 others (auth and identity products).
- [Terms allow an agent's account](https://agentdrives.com/standard#in-10): **The same as all** 8 others (auth and identity products).
- [Errors say what to do next](https://agentdrives.com/standard#err-03): **The same as all** 6 others (auth and identity products).
- [WebMCP tools on the web app](https://agentdrives.com/standard#proto-02): **The same as all** 5 others (auth and identity products).
- [Spend an agent can see and cap](https://agentdrives.com/standard#pay-05): **Level with 6 of 7** others (auth and identity products), 1 higher.
- [Agents pick the right tool](https://agentdrives.com/standard#mcp-05): **Level with 72 of 91** others (every product), 19 higher.
- [Every way in works for an agent](https://agentdrives.com/standard#in-11): **Level with 5 of 7** others (auth and identity products), 2 higher.
- [Delete by API, checked](https://agentdrives.com/standard#leave-02): **Level with 4 of 7** others (auth and identity products), 3 higher.
- [Limits known before they hit](https://agentdrives.com/standard#lim-02): **Tied for the lowest result** 2 of 4 others (auth and identity products) the same, 2 higher.
- [Versions an agent can pin](https://agentdrives.com/standard#api-07): **Behind 3 of 5** others (auth and identity products), 2 the same.
- [Truthful annotations](https://agentdrives.com/standard#mcp-07): **Behind 68 of 91** others (every product), 23 the same.
- [Descriptions that say when to use the tool](https://agentdrives.com/standard#mcp-04): **Behind 70 of 91** others (every product), 21 the same.
- [Typed, tight input schemas](https://agentdrives.com/standard#mcp-06): **Behind 75 of 91** others (every product), 16 the same.
- [Prices an agent can read](https://agentdrives.com/standard#find-06): **Behind 6 of 7** others (auth and identity products), 1 the same.
- [Export before leaving](https://agentdrives.com/standard#leave-03): **Behind 6 of 7** others (auth and identity products), 1 the same.
- [Docs an agent can navigate](https://agentdrives.com/standard#docs-11): **The lowest result** All 7 others (auth and identity products) higher.
- [Real status codes](https://agentdrives.com/standard#err-01): **The lowest result** All 7 others (auth and identity products) higher.

[How each criterion is graded](https://agentdrives.com/standard)

## Agents get stuck at Leave

**6** things to fix for agents: 4 small and 2 medium

One fix can cover several findings.

Agent blockers: **4 found** Agent surfaces: **26 issues found**

[Get your findings, free](https://agentdrives.com/your-findings?p=humanmirror) · [What a drive covers](https://agentdrives.com/method?via=md)

We mail a private link to an address at humanmirror.fr.

1. Find Partly
2. Get in Clear
3. Use Partly
4. Pay Clear
5. Leave Blocks agents

Or [claim HumanMirror](https://agentdrives.com/go/claim/humanmirror?add=humanmirror.fr&ref=listing&safety=1) to see where each issue is, read your safety findings with their fixes and ask for a free drive after a fix.

Want HumanMirror driven every month? [See Watch](https://agentdrives.com/owners#watch)

### On our standard

Best or ahead on 11 of 33 criteria, behind on 9. Each against the products we measured on it.

- [Pages fit one fetch](https://agentdrives.com/standard#docs-08)
- [An agent payment rail on the core](https://agentdrives.com/standard#pay-02)
- [A plan an agent can buy by API](https://agentdrives.com/standard#pay-03)
- [Structured results](https://agentdrives.com/standard#mcp-08)
- [Prices an agent can read](https://agentdrives.com/standard#find-06)
- [Export before leaving](https://agentdrives.com/standard#leave-03)
- [Docs an agent can navigate](https://agentdrives.com/standard#docs-11)
- [Real status codes](https://agentdrives.com/standard#err-01)

**All 33 criteria**

- [A request id on every answer](https://agentdrives.com/standard#err-07)
- [A path written for agents](https://agentdrives.com/standard#docs-10)
- [A person can claim what the agent made](https://agentdrives.com/standard#in-09)
- [A tool list that fits](https://agentdrives.com/standard#mcp-11)
- [Answers sized for an agent](https://agentdrives.com/standard#api-04)
- [Errors a program can read](https://agentdrives.com/standard#err-02)
- [The best way in is built for agents](https://agentdrives.com/standard#in-12)
- [Results sized for a context window](https://agentdrives.com/standard#mcp-09)
- [Retried payments charge once, with a receipt](https://agentdrives.com/standard#pay-04)
- [A way in with no person](https://agentdrives.com/standard#in-01)
- [Tool names that sort themselves](https://agentdrives.com/standard#mcp-03)
- [llms.txt that works](https://agentdrives.com/standard#find-02)
- [Fast to the first call](https://agentdrives.com/standard#in-03)
- [Terms allow an agent's account](https://agentdrives.com/standard#in-10)
- [Errors say what to do next](https://agentdrives.com/standard#err-03)
- [WebMCP tools on the web app](https://agentdrives.com/standard#proto-02)
- [Spend an agent can see and cap](https://agentdrives.com/standard#pay-05)
- [Agents pick the right tool](https://agentdrives.com/standard#mcp-05)
- [Every way in works for an agent](https://agentdrives.com/standard#in-11)
- [Delete by API, checked](https://agentdrives.com/standard#leave-02)
- [Limits known before they hit](https://agentdrives.com/standard#lim-02)
- [Versions an agent can pin](https://agentdrives.com/standard#api-07)
- [Truthful annotations](https://agentdrives.com/standard#mcp-07)
- [Descriptions that say when to use the tool](https://agentdrives.com/standard#mcp-04)
- [Typed, tight input schemas](https://agentdrives.com/standard#mcp-06)

[How each criterion is graded](https://agentdrives.com/standard)

## For the team behind HumanMirror

**6 things to fix for agents: 4 small and 2 medium.**

One fix can cover several findings. Agent blockers: 4 found · Agent surfaces: 26 issues found.

Agents get stuck at Leave. Its team gets the two findings that matter most in full, free, by a private link to an address at humanmirror.fr.

- Find: Partly
- Get in: Clear
- Use: Partly
- Pay: Clear
- Leave: Blocks agents

[Get your findings, free](https://agentdrives.com/your-findings?p=humanmirror) · [Claim HumanMirror](https://agentdrives.com/app/products?add=humanmirror.fr&ref=listing&safety=1) · [What a drive covers](https://agentdrives.com/method?via=md)

## Read more on this step

Leave is where our agent stopped at HumanMirror.

- [Agents can't leave](https://agentdrives.com/findings/no-way-out?via=md) (finding): 43 of the 96 products where our agent got as far as deleting its account gave it no way to do it on its own.
- [Let AI agents delete the account they opened](https://agentdrives.com/guides/let-agents-delete-their-account?via=md) (guide): A delete call for the account, a clear answer when it's done, and the same in the settings for people.
- [Agent experience (AX): what it is, measured at 246 products](https://agentdrives.com/guides/agent-experience?via=md) (guide): As of 9 Nov 2026: Agent experience (AX) is whether an AI agent can find, get into, use and pay for a product on its own. We drove 246 products to measure it.

[Everything on Leave](https://agentdrives.com/access/leave)

## Questions people ask about HumanMirror

**Can an AI agent create a HumanMirror account on its own?**

Yes, over the API, without a browser. OmniDome agent registration and a Forge trial key also without a browser.

**Does HumanMirror have an MCP server, and can an agent connect to it?**

Yes. HumanMirror has an official MCP server. A new agent can get in on its own: in the drive of 15 Nov 2026, our agent signed up and got its key without a person. Through the MCP server, our agent tried 3 tasks and did 2.

**How does an AI agent get a HumanMirror API key?**

In the drive of 15 Nov 2026, our agent got its HumanMirror API key over the API, without a browser or a person. The trial POST returns an hm_nexus_ Bearer key in its 201 body and it works at once; no route to create a scoped key or to revoke or rotate one in any spec.

**Can an AI agent delete its HumanMirror account?**

Not on its own: blocked. No delete or revoke route was found in the 8 OpenAPI specs, the MCP tools or llms.txt; DELETE on the Nexus and OmniDome endpoints answered 405. The privacy page sends erasure requests by mail to privacy@humanmirror.fr, which our agent may not write to. Still there: the Nexus trial key, the Forge trial key, the OmniDome agent identity and key, and the OmniDome child agent.

**Can an AI agent pay for HumanMirror?**

Yes. Valid x402 payment challenge: $0.001 a call.

## Where HumanMirror fits

- Category: [Developer platform](https://agentdrives.com/categories/developer?via=md), [Auth and identity](https://agentdrives.com/categories/auth?via=md)
- Tasks: [Read usage and quotas](https://agentdrives.com/tasks/read-usage-and-quotas), [Search available apps and actions](https://agentdrives.com/tasks/search-available-apps-and-actions)
- Standards: [MCP](https://agentdrives.com/standards/mcp) (tools an agent can call), [OpenAPI](https://agentdrives.com/standards/openapi) (its API, described), [llms.txt](https://agentdrives.com/standards/llms-txt) (its docs, listed for agents), [x402](https://agentdrives.com/standards/x402) (pay per request)
- Steps: [Find](https://agentdrives.com/access/find), [Get in](https://agentdrives.com/access/get-in), [Use](https://agentdrives.com/access/use), [Pay](https://agentdrives.com/access/pay), [Leave](https://agentdrives.com/access/leave) (first stop)
- Topics: [MCP servers](https://agentdrives.com/access/find/mcp-servers) (Find step), [Docs and llms.txt](https://agentdrives.com/access/find/docs) (Find step), [MCP sign-in](https://agentdrives.com/access/get-in/mcp-sign-in) (Get in step), [Sign-up](https://agentdrives.com/access/get-in/sign-up) (Get in step), [MCP and API](https://agentdrives.com/access/use/mcp-and-api) (Use step), [Card checkout](https://agentdrives.com/access/pay/card-checkout) (Pay step), [Delete the account](https://agentdrives.com/access/leave/delete-account) (Leave step)
- Rivals: [Auth0](https://agentdrives.com/tools/auth0?via=md) (L3), [Clerk](https://agentdrives.com/tools/clerk?via=md) (L3), [Kinde](https://agentdrives.com/tools/kinde?via=md) (L3)

The level follows [these rules](https://agentdrives.com/changes#rules-2026-11-18).

## Level by drive

One drive so far, on 15 Nov 2026.

[Every drive in full](https://agentdrives.com/tools/humanmirror/history?via=md)

Source: our drives of HumanMirror. As of 15 Nov 2026.

One drive so far.

Data: [JSON](https://agentdrives.com/api/tools/humanmirror.json?via=md) · [Markdown](https://agentdrives.com/tools/humanmirror.md?via=md) · [History](https://agentdrives.com/tools/humanmirror/history?via=md) · [Atom](https://agentdrives.com/feeds/tools/humanmirror.atom) · [JSON Feed](https://agentdrives.com/feeds/tools/humanmirror.json) · licence [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) · evidence sha256 `717f7f53a7b87bd5a80a3618cf0f7842bc4c8bfcf4d03df0274462852127533b`

**Found an error?** [Tell us on this page](https://agentdrives.com/tools/humanmirror?via=md#correct) or read [how corrections work](https://agentdrives.com/corrections?via=md).
