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

Mercury is a business bank with a test sandbox.

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

Rank among the 246 products we drove:

- Core work done: 88%, tied #60 (7 of 8 core tasks)
- Steps on its own: 3 / 5, tied #35 (Find to Leave, no person)
- Time to get in: 8 minutes, tied #83 (Getting in on its own)
- Holds up: 67 / 100, tied #92 (Reliability under errors)

**Can an AI agent use Mercury? Yes.** Our agent signed up for Mercury on its own in a browser with an email and password, got an API key and used its API and MCP server. No CLI found.

- Level confirmed: Not yet (Level from one drive; a second must agree)
- Got in by: Sign-up in a browser (8 minutes)
- Got as far as: Pay: ready to pay (Payment not yet proven; stopped at the pay button of the $35 a month charge)

Category: Banking APIs · Connects by: API · Agent payments: None found · Driven: 2 Nov 2026

**How to connect**
- MCP server URL: `https://mcp.mercury.com/mcp` (documented). Source: https://mcp.mercury.com/mcp
- Docs page: `https://docs.mercury.com/reference/getaccountcards` (tested in our drive). Source: https://docs.mercury.com/reference/getaccountcards

Checked 18 Nov 2026: MCP server asks for a sign-in; sign-up page unchanged since the check before.

See [how it ranks in the banking APIs category](https://agentdrives.com/categories/banking-apis?via=md).

## From finding it to leaving

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

1. **Find**: On its own. Docs it can act on: llms.txt. (148 doc pages read)
2. **Get in**: On its own. No key needed, MCP sign-in, sign-up in a browser. (8 minutes)
3. **Use**: On its own. Did: send payments and read them back; add, edit and delete payment recipients. (7 of 8 core tasks)
4. **Pay**: Partly. Ready to pay. Payment not yet proven; stopped at the pay button of the $35 a month charge. First stop.
5. **Leave**: Partly. Nothing to cancel. Its account stays.

To reach L4: Delete its account, so its keys and login stop working.

## How an agent connects

API and llms.txt found.

- **MCP server**: 10 of 42 tested. A server for searching the docs only. (42 tools mapped in our drive)
- **API**: 75 of 82 tested. REST, OpenAPI embedded per reference page. (82 operations mapped in our drive)
- **OpenAPI**: Unknown.
- **CLI**: None found.
- **SDK**: None found.
- **llms.txt**: Yes. https://docs.mercury.com/llms.txt
- **Agent payments**: None found. Walked to the last click ("Start free trial") and stopped: Mercury Books, $35 a month. It checks out through its own checkout. No CAPTCHA at checkout.
- **Docs**: In part. Covers getting started, auth, error codes and code examples. Describes each operation only in part: completeness 50 of 100. 9 of 10 documented ways to do the core work were tried, and 6 of them worked. https://docs.mercury.com/reference/getaccountcards (148 of 150 pages read)

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

## Getting in

4 of 4 ways in worked without a person.

The way that worked: **Sign-up in a browser**, 8 minutes. Needs an inbox and a browser. No person needed.

- Agent on a server (API, MCP, CLI): Partly. Gets in without a browser (no key needed) and does 0 of its 8 core tasks that way. For a full account, it stops at sign-up: it works only in a browser.
- Assistant in its own browser (Like ChatGPT agent): Partly. Signs up, confirms its email and does 7 of its 8 core tasks, then stops: its account stays. It can't delete its account.
- Signed agent (Known to Cloudflare): Yes. The sign-up page opens.

| Way in | What happened |
|---|---|
| Sign-up in a browser | Worked |
| Sign-up in a browser | Worked |
| MCP sign-in | Worked |
| No key needed | Worked: Part of the work done |

- Crawlers: **The robots.txt on sandbox.mercury.com asks crawlers to stay off sign-up.** Nothing to fix.

## What an agent can do with Mercury

7 of 8 core tasks done, and 16 of 27 others.

- Holds up: 67 of 100 (Reliability 87, Handles mistakes 47). The mean of the two.
- Docs match: 81 of 100 (Docs accuracy 94, Docs completeness 50). Docs accuracy counts 70%, completeness 30%.
- Surface tested: 63 of 100. 84 of 133 passed: the 136 documented operations less the 3 that spend money or reach real people

Core tasks:

| Task | Result | Done through | Tried, not done | Offered, not tried |
|---|---|---|---|---|
| Send money to a recipient | Done | API | – | – |
| Add a payment recipient | Done | API | – | – |
| Transfer money between own accounts | Done | API | – | MCP |
| Read the account balance | Done | API | – | MCP |
| List and search account transactions | Done | API | – | MCP |
| Create and send an invoice | Done | API | – | – |
| Pay a bill | Done | Web app | – | – |
| Issue a virtual card | Blocked: The API returns 400 or 500 and the web app shows an error. | – | API | – |

16 of 27 other tasks done:

- Change a card's nickname or spending limit: Done
- Freeze or unfreeze a card: Done
- Categorize a transaction or add a note: Done
- Attach a receipt to a transaction: Done
- Create a custom expense category: Done
- Edit or cancel an invoice: Done
- Set up recurring invoices: Done
- Invite a recipient to submit payment details: Done
- Create a spending budget: Done
- Ask Command to do financial work: Done
- View spend insights by category: Done
- Keep the books in Mercury Books: Done
- Register a webhook: Done
- Read the event stream of changes: Done
- Invite a team member: Done
- Restrict API token access by IP address: Done
- Download an account statement: Partly
- Move cash into Treasury: Partly
- Create a SAFE: Partly
- Request a payment that needs approval: Blocked
- Approve a pending payment: Blocked
- Run payroll: Blocked
- Cancel a card: Not tried
- Create an invoice customer: Not tried
- Apply for a working capital loan: Not tried
- Set approval rules and user permissions: Not tried
- Connect QuickBooks, Xero or NetSuite: Not tried

What it can't do, and why:

- Issue a virtual card in the sandbox: The API returns 400 or 500 and the web app shows an error.
- Create or revoke an API token without the dashboard: No API, CLI or device login is offered for tokens.
- Get a payment approved: A second approver needs a phone number; the requester cannot approve.
- Use the Mercury MCP server with sandbox data: MCP only signs in production accounts, which need a phone and EIN.
- Close the account: Closing needs a zero balance, and the login stays afterwards.

**Surface audit:** 136 documented operations: 89 run, 44 blocked by sandbox or plan limits and 3 left out because they spend money or reach real people. Of the 89 run, 84 passed and 5 didn't; 5 behaved differently from the docs. By surface, 75 on the API, 10 on the MCP server and 4 elsewhere, such as sign-in and webhooks.

## Tips for agent builders

- Use sandbox.mercury.com and the API base https://api-sandbox.mercury.com/api/v1/; production tokens do not work there.
- At the 2FA step choose "Other authentication app" and keep the shown key to compute TOTP codes yourself.
- Put the IPv4 your HTTP client really uses on a Read and Write token; the ipNotWhitelisted 401 names it.
- To pay a bill, create it in Books > Bills, pay the vendor with POST /account/{id}/transactions, then Match bill on that transaction.
- Do not repeat a payment to the same recipient for the same amount within 24 hours; it is refused as a duplicate even with a new idempotency key.
- Use the docs MCP at https://docs.mercury.com/mcp (get-endpoint) to read the exact request schema for any operation.

## Paying

Ways to pay checked 2 Nov 2026.

- Result: Partly. Ready to pay. Payment not yet proven; stopped at the pay button of the $35 a month charge.
- Checkout: Walked to the last click ("Start free trial") and stopped: Mercury Books, $35 a month. It checks out through its own checkout. No CAPTCHA at checkout.
- Ways to pay:
  - Stripe card form (card checkout)
- Trial: It has a free plan.

## Leaving

- Result: Partly. Nothing to cancel. Its account stays.
- Its account: Not on its own: blocked. The sandbox account and the production application both stay: close is a request flow blocked until the balance is zero and pending items settle, the login stays after closing and data is kept 7 years; production has no close option. API data, token and webhook were removed.

## 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; 14 of 14 days
- llms.txt, not on its site on the newest check (its drive found one elsewhere); 14 of 14 days
- MCP server, asks for a sign-in on the newest check; 14 of 14 days
- OpenAPI, not on its site on the newest check; 14 of 14 days

## How Mercury compares

On our standard (31 criteria, each against every product we measured on it). Best or ahead on 5 of 31 criteria, behind on 4.

- [Every way in works for an agent](https://agentdrives.com/standard#in-11): **Ahead of 134 of 193** others, 59 the same.
- [Docs an agent can navigate](https://agentdrives.com/standard#docs-11): **Ahead of 128 of 187** others, 59 the same.
- [Spend an agent can see and cap](https://agentdrives.com/standard#pay-05): **Ahead of 106 of 159** others, 46 the same, 7 higher.
- [A way in with no person](https://agentdrives.com/standard#in-01): **Ahead of 137 of 243** others, 106 the same.
- [Truthful annotations](https://agentdrives.com/standard#mcp-07): **Ahead of 50 of 91** others, 41 the same.
- [Prices an agent can read](https://agentdrives.com/standard#find-06): **Level with 117 of 153** others, 36 lower.
- [Errors a program can read](https://agentdrives.com/standard#err-02): **Level with 82 of 145** others, 47 lower, 16 higher.
- [Agents may read docs, API and sign-up](https://agentdrives.com/standard#find-01): **Level with 129 of 160** others, 31 lower.
- [Fast to the first call](https://agentdrives.com/standard#in-03): **Level with 118 of 135** others, 17 lower.
- [Descriptions that say when to use the tool](https://agentdrives.com/standard#mcp-04): **Level with 56 of 91** others, 22 lower, 13 higher.
- [The best way in is built for agents](https://agentdrives.com/standard#in-12): **In the middle** 79 of 196 others lower, 55 the same, 62 higher.
- [Tool names that sort themselves](https://agentdrives.com/standard#mcp-03): **Level with 85 of 91** others, 6 lower.
- [Terms allow an agent's account](https://agentdrives.com/standard#in-10): **Level with 179 of 189** others, 10 lower.
- [Real status codes](https://agentdrives.com/standard#err-01): **Level with 161 of 168** others, 7 lower.
- [Errors say what to do next](https://agentdrives.com/standard#err-03): **Level with 107 of 145** others, 22 lower, 16 higher.
- [Typed, tight input schemas](https://agentdrives.com/standard#mcp-06): **Level with 59 of 91** others, 17 lower, 15 higher.
- [llms.txt that works](https://agentdrives.com/standard#find-02): **Level with 127 of 239** others, 57 lower, 55 higher.
- [A path written for agents](https://agentdrives.com/standard#docs-10): **In the middle** 48 of 187 others lower, 91 the same, 48 higher.
- [WebMCP tools on the web app](https://agentdrives.com/standard#proto-02): **Level with 133 of 136** others, 3 higher.
- [An agent payment rail on the core](https://agentdrives.com/standard#pay-02): **Level with 205 of 216** others, 11 higher.
- [A request id on every answer](https://agentdrives.com/standard#err-07): **In the middle** 42 of 143 others lower, 47 the same, 54 higher.
- [Results sized for a context window](https://agentdrives.com/standard#mcp-09): **In the middle** 16 of 70 others lower, 27 the same, 27 higher.
- [Agents pick the right tool](https://agentdrives.com/standard#mcp-05): **Level with 72 of 91** others, 19 higher.
- [A tool list that fits](https://agentdrives.com/standard#mcp-11): **In the middle** 21 of 91 others lower, 26 the same, 44 higher.
- [Pages fit one fetch](https://agentdrives.com/standard#docs-08): **Level with 118 of 187** others, 69 higher.
- [All validation problems at once](https://agentdrives.com/standard#err-04): **Level with 33 of 57** others, 24 higher.
- [Limits known before they hit](https://agentdrives.com/standard#lim-02): **Level with 47 of 90** others, 43 higher.
- [Sign-in for agents acting for a person](https://agentdrives.com/standard#in-08): **Behind 84 of 158** others, 74 the same.
- [Delete by API, checked](https://agentdrives.com/standard#leave-02): **Behind 52 of 97** others, 45 the same.
- [Export before leaving](https://agentdrives.com/standard#leave-03): **Behind 117 of 189** others, 72 the same.
- [Structured results](https://agentdrives.com/standard#mcp-08): **Behind 57 of 78** others, 21 the same.

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

## Agents get stuck at Pay

**9** things to fix for agents: 4 small, 4 medium and 1 large

One fix can cover several findings.

Agent blockers: **5 found** Agent surfaces: **5 issues found**

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

We mail a private link to an address at mercury.com.

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

Or [claim Mercury](https://agentdrives.com/go/claim/mercury?add=mercury.com&ref=listing) to see where each issue is and ask for a free drive after a fix.

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

### On our standard

Best or ahead on 5 of 31 criteria, behind on 4. Each against every product we measured on it.

- [Every way in works for an agent](https://agentdrives.com/standard#in-11)
- [Docs an agent can navigate](https://agentdrives.com/standard#docs-11)
- [Spend an agent can see and cap](https://agentdrives.com/standard#pay-05)
- [A way in with no person](https://agentdrives.com/standard#in-01)
- [Sign-in for agents acting for a person](https://agentdrives.com/standard#in-08)
- [Delete by API, checked](https://agentdrives.com/standard#leave-02)
- [Export before leaving](https://agentdrives.com/standard#leave-03)
- [Structured results](https://agentdrives.com/standard#mcp-08)

**All 31 criteria**

- [Truthful annotations](https://agentdrives.com/standard#mcp-07)
- [Prices an agent can read](https://agentdrives.com/standard#find-06)
- [Errors a program can read](https://agentdrives.com/standard#err-02)
- [Agents may read docs, API and sign-up](https://agentdrives.com/standard#find-01)
- [Fast to the first call](https://agentdrives.com/standard#in-03)
- [Descriptions that say when to use the tool](https://agentdrives.com/standard#mcp-04)
- [The best way in is built for agents](https://agentdrives.com/standard#in-12)
- [Tool names that sort themselves](https://agentdrives.com/standard#mcp-03)
- [Terms allow an agent's account](https://agentdrives.com/standard#in-10)
- [Real status codes](https://agentdrives.com/standard#err-01)
- [Errors say what to do next](https://agentdrives.com/standard#err-03)
- [Typed, tight input schemas](https://agentdrives.com/standard#mcp-06)
- [llms.txt that works](https://agentdrives.com/standard#find-02)
- [A path written for agents](https://agentdrives.com/standard#docs-10)
- [WebMCP tools on the web app](https://agentdrives.com/standard#proto-02)
- [An agent payment rail on the core](https://agentdrives.com/standard#pay-02)
- [A request id on every answer](https://agentdrives.com/standard#err-07)
- [Results sized for a context window](https://agentdrives.com/standard#mcp-09)
- [Agents pick the right tool](https://agentdrives.com/standard#mcp-05)
- [A tool list that fits](https://agentdrives.com/standard#mcp-11)
- [Pages fit one fetch](https://agentdrives.com/standard#docs-08)
- [All validation problems at once](https://agentdrives.com/standard#err-04)
- [Limits known before they hit](https://agentdrives.com/standard#lim-02)

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

## For the team behind Mercury

**9 things to fix for agents: 4 small, 4 medium and 1 large.**

One fix can cover several findings. Agent blockers: 5 found · Agent surfaces: 5 issues found.

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

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

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

## Read more on this step

Pay is where our agent stopped at Mercury.

- [How far does an AI agent get in a card checkout?](https://agentdrives.com/findings/card-checkout?via=md) (finding): Our agent reached the checkout at 88 of the 190 products it set out to buy from, and the card form at 68.
- [Let AI agents pay you per call: x402 and MPP](https://agentdrives.com/guides/let-agents-pay-per-call?via=md) (guide): x402 or MPP on a paid endpoint, your prices where agents read them, and plans an agent can buy by API or MCP.
- [Tell an AI agent a feature needs a paid plan](https://agentdrives.com/guides/tell-ai-agents-a-feature-needs-a-paid-plan?via=md) (guide): How to answer an AI agent at a paid-plan wall: 402 or 403, a Problem Details body with the required plan, and an upgrade link a person can open.

[Everything on Pay](https://agentdrives.com/access/pay)

## Questions people ask about Mercury

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

Yes, in a browser, without a person. Sandbox sign-up in the browser (name, work email, password), then mandatory TOTP 2FA the agent set up itself and backup codes; no sign-up API. Production sign-up separately walled at phone number and EIN.

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

Mercury's docs have an MCP server for searching them. The drive found no MCP server that works the product itself. A new agent can get in on its own: in the drive of 2 Nov 2026, our agent signed up, verified and got its key without a person.

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

In the drive of 2 Nov 2026, our agent got its Mercury API key in a browser, without a person. Token made by the agent in the sandbox dashboard (Settings > Tokens) with Read and Write scope and a required IP whitelist; no API or CLI route to create or revoke tokens exists. Revoked in the dashboard, after which the token gets 401.

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

Not on its own: blocked. The sandbox account and the production application both stay: close is a request flow blocked until the balance is zero and pending items settle, the login stays after closing and data is kept 7 years; production has no close option. API data, token and webhook were removed.

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

Partly. Ready to pay: payment not yet proven; stopped at the pay button of the $35 a month charge. It has a free plan.

## Where Mercury fits

- Category: [Payments and finance](https://agentdrives.com/categories/finance?via=md), [Banking APIs](https://agentdrives.com/categories/banking-apis?via=md)
- Tasks: [Register a webhook](https://agentdrives.com/tasks/register-a-webhook), [Invite a member](https://agentdrives.com/tasks/invite-a-member), [Read the audit log](https://agentdrives.com/tasks/read-the-audit-log), [Restrict access by IP address](https://agentdrives.com/tasks/restrict-access-by-ip-address), [Delegate a task to the built-in AI agent](https://agentdrives.com/tasks/delegate-a-task-to-the-built-in-ai-agent), [Read the account balance](https://agentdrives.com/tasks/read-the-account-balance)
- Standards: [llms.txt](https://agentdrives.com/standards/llms-txt) (its docs, listed for agents)
- 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) (first stop), [Leave](https://agentdrives.com/access/leave)
- Topics: [Docs and llms.txt](https://agentdrives.com/access/find/docs) (Find step), [Sign-up](https://agentdrives.com/access/get-in/sign-up) (Get in step), [Card checkout](https://agentdrives.com/access/pay/card-checkout) (Pay step), [Delete the account](https://agentdrives.com/access/leave/delete-account) (Leave step)

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

## Level by drive

One drive so far, on 2 Nov 2026.

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

Source: our drives of Mercury. As of 2 Nov 2026.

1. 14 Nov 2026 [**Where it stopped changed: Pay → Delete**](https://agentdrives.com/tools/mercury?via=md#drive-20261102T182906Z) [Re-scored under method v49, drive of 2 Nov 2026](https://agentdrives.com/changes#rules-2026-11-12)
2. 11 Nov 2026 [**Level L4 → L3**](https://agentdrives.com/tools/mercury?via=md#drive-20261102T182906Z) [Re-scored under method v47, drive of 2 Nov 2026](https://agentdrives.com/changes#rules-2026-11-11-47)

Data: [JSON](https://agentdrives.com/api/tools/mercury.json?via=md) · [Markdown](https://agentdrives.com/tools/mercury.md?via=md) · [History](https://agentdrives.com/tools/mercury/history?via=md) · [Atom](https://agentdrives.com/feeds/tools/mercury.atom) · [JSON Feed](https://agentdrives.com/feeds/tools/mercury.json) · licence [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) · evidence sha256 `ff5485e01ccc2e1786df83c521f6eee478fd21e265eb1189237aef7503331263`

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