Plunk
useplunk.comPlunk is an email service for sending and automating email.
Can an AI agent use Plunk? Yes, without signing up. Our agent got into Plunk on its own in 5 minutes and did 2 of its 7 core tasks.
From finding it to leaving
What an agent managed without a person, at each step.
- On its own
Find
Docs it can act on: llms.txt, an OpenAPI and an MCP server.
176 doc pages read
- On its own
Get in
Runs locally, key by API, sign-up in a browser.
5 minutes
- Partly
Use
Did: manage contacts, custom fields and CSV imports; build segments and track events. Not done: send any email, template or campaign.
2 of 7 core tasks
- Reaches the checkout
Pay
Stopped at the card form.
- On its own
Leave
Nothing to cancel, deletes its account.
Getting in
3 of 3 ways in worked without a person.
The way that worked
Runs locally, no account5 minutes
- An inboxNot needed
- A browserNot needed
- A personNot needed
- Agent on a serverAPI, MCP, CLIPartlySigns up, confirms its email, gets an API key, makes its first API call and does 2 of its 7 core tasks, then stops: the rest needs a paid plan.
- Assistant in its own browserLike ChatGPT agentPartlySigns up, confirms its email and does 2 of its 7 core tasks, then stops: the rest needs a paid plan.
- Signed agentKnown to CloudflareYesThe sign-up page opens.
Every way in
| Way in | What happened |
|---|---|
| Runs locally, no account | Worked |
| Sign-up by API | Worked |
| Email and password | Worked |
What an agent can do with Plunk
2 of 7 core tasks done, and 11 of 19 others.
Core tasks
- done there
- tried, not done
| Task | Result | Through |
|---|---|---|
| Build an email automation | Done | API: done thereMCP: done there |
| Add a contact to an automation | Done | API: done thereMCP: done there |
| Send a transactional email | Blocked | API: tried, not done |
| Send an email from a templateEvery send needs a sender domain verified by DNS records. | Blocked | API: tried, not done |
| Draft an email campaign | Blocked | API: tried, not done |
| Schedule an email campaign | Blocked | API: tried, not done |
| Receive inbound emailNeeds a verified domain and an MX record. | Blocked | API: tried, not done |
Reliability
92of 100
Core tasks that worked on all 3 runs
Handles mistakes
65of 100
Clear errors, retries without duplicates, honest limits
Surface tested
77of 100
Documented routes, tools and commands run and checked
Docs accuracy
87of 100
Documented operations that behaved as the docs say
Docs completeness
69of 100
How fully the docs describe each operation
What it can't do, and why
- Send any email, template or campaign Every send needs a sender domain verified by DNS records.
- Receive inbound email Needs a verified domain and an MX record.
- Pay for the paid plan Only card checkout exists.
- Delete the account through documented API The account is deleted in the dashboard only.
11 of 19 other tasks done
- Validate an email addressDone
- Create a list and add contacts to itDone
- Update a contact or subscriberDone
- Delete a contact or subscriberDone
- Import contacts from a CSV fileDone
- Track an event for a contactDone
- Create a custom contact fieldDone
- Create a segmentDone
- Manage suppressions (bounce and complaint handling)Done
- Create an API keyDone
- Read usage and quotasDone
- Add and verify a sending domainPartly
- Track email opens and clicksPartly
- Register a webhookPartly
- Translate the unsubscribe footer and contact-facing pagesPartly
- Self-host the platform on your own infrastructurePartly
- Remove the product's branding from emailsPaid plan
- Set a spend limitPaid plan
- Create an email template: Every send needs a sender domain verified by DNS records.Blocked
- Read campaign performanceBlocked
- Set up a preference centre or unsubscribe pageBlocked
Paid-plan tasks don't count toward the level. The drive used the free plan.
Surface audit: 169 documented operations, 133 run, 123 passed, 17 not as documented, 27 blocked (sandbox or plan limits), 9 left out to stay safe.
How an agent connects
MCP, API, OpenAPI and llms.txt found.
MCP server
37 testedOfficial server. Stdio server @plunk/mcp 0.17.0, 42 tools.
45 tools across its docs
API
90 testedREST at next-api.useplunk.com, about 115 documented endpoints.
117 operations across its docs
OpenAPI
Yes3.1.0, 18 operations; workflows, domains and events are not in it.
CLI
None foundSDK
None foundllms.txt
YesAgent payments
None foundThe checkout: Reached the card form and stopped: Pay-as-you-go; Checkout: Stripe Checkout; An invisible bot check loaded at checkout; Wallets: Link.
Docs
No gaps in what we readCovers getting started, auth, error codes and code examples.
2 of 6 documented ways to do the core work worked when tried.
176 of 177 pages read
Tips for agent builders
- Use sk_ for everything except POST /v1/track, which takes pk_.
- Check GET /domains/project/:projectId (or plunk_list_domains) before sending: without a verified domain every send, template and campaign returns 403.
- Workflow steps need a 'position'; UPDATE_CONTACT takes config.updates. Read the execution back: a bad config only fails at run time.
- Prefer the MCP server's plunk_add_workflow_steps and plunk_check_workflow: they validate before saving.
- Bulk contact jobs take {"mode":"ids","contactIds":[...]}.
How Plunk compares
In Email infrastructure
2nd of 2 Email infrastructure products our agent got into on its own.
| Product | Level | Find to leave |
|---|---|---|
| L5 | Find: On its own, Get in: On its own, Use: On its own, Pay: On its own, Leave: On its own. | |
| Plunk (this page) | L2 | Find: On its own, Get in: On its own, Use: Partly, Pay: Partly, Leave: On its own. |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Partly, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. |
Sorted by level, then work done, then time to the first call. How we rank
For the team behind Plunk
Our agent got through Use and Pay only in part. Prove Plunk is yours, then see where each issue is, read your safety findings and ask for a free drive after a fix. How to fix each one comes with the Audit.
11 things to fix for agents
Agent blockers: 1 foundAgent surfaces: 18 issues foundSafety review: no issues found
- FindClear
- Get inClear
- UsePartly
- PayPartly
- LeaveClear
On our standard
18 criteria: against other email infrastructure products where we measured 5 or more, otherwise against every product we measured.
- Every way in works for an agentThe best result All 5 others (email infrastructure products) lower.
- The best way in is built for agentsAhead of most 4 of 5 others (email infrastructure products) lower, 1 the same.
- Pages fit one fetchAhead of most 3 of 5 others (email infrastructure products) lower, 2 the same.
- Docs an agent can navigateAhead of most 3 of 5 others (email infrastructure products) lower, 2 the same.
- Export before leavingLevel with most 3 of 5 others (email infrastructure products) the same, 2 higher.
- A person can claim what the agent madeBehind most 6 of 9 others (every product) higher, 3 the same.
- Versions an agent can pinBehind most 3 of 4 others (email infrastructure products) higher, 1 the same.
- Real status codesBehind most 126 of 127 others (every product) higher, 1 the same.
All 18 criteria
- One way to page through listsIn the middle 35 of 87 others (every product) lower, 42 the same, 10 higher.
- Answers sized for an agentLevel with most 61 of 85 others (every product) the same, 24 lower.
- Limits known before they hitAhead of most 27 of 47 others (every product) lower, 5 the same, 15 higher.
- Errors a program can readIn the middle 40 of 104 others (every product) lower, 43 the same, 21 higher.
- Fast to the first callLevel with most 71 of 85 others (every product) the same, 14 lower.
- Errors say what to do nextLevel with most 56 of 103 others (every product) the same, 24 lower, 23 higher.
- Prices an agent can readThe same as all 5 others (email infrastructure products).
- A path written for agentsLevel with most 3 of 5 others (email infrastructure products) the same, 1 lower, 1 higher.
- WebMCP tools on the web appLevel with most 91 of 93 others (every product) the same, 2 higher.
- A request id on every answerIn the middle 34 of 103 others (every product) lower, 29 the same, 40 higher.
Where Plunk fits
Cited in 2 articles
Found an error? Tell us on this page or read how corrections work.
Questions people ask about Plunk
- Can an AI agent sign up for Plunk on its own?
- Yes, over the API, without a browser.
- Does Plunk have an MCP server, and can an agent connect to it?
- Yes. Plunk has an official MCP server. The server runs on the agent's own machine and lists 42 tools. A new agent can get in on its own: in the drive of 27 Oct 2026, our agent signed up, verified and got its key without a person.
- How does an AI agent get a Plunk API key?
- In the drive of 27 Oct 2026, our agent got its Plunk API key over the API, without a browser or a person.
- Can an AI agent delete its Plunk account?
- Yes, in a browser, without a human. Project deleted without a browser; the account only in the dashboard (Account → Delete account after deleting the project). Afterwards login, session and secret key are refused. An uploaded image stays public.
- Can an AI agent pay for Plunk?
- Partly. Reaches the checkout: stopped at the card form. No card needed to start.
Cite this page
The page as a reference, in plain text, APA and BibTeX.
Cite this
AgentDrives, “Plunk: AI agent test drive”, 27 October 2026, https://agentdrives.com/tools/useplunk. CC BY 4.0.
AgentDrives. (2026, October 27). Plunk: AI agent test drive. https://agentdrives.com/tools/useplunk
@misc{agentdrives_useplunk_2026,
author = {{AgentDrives}},
title = {Plunk: AI agent test drive},
year = {2026},
month = oct,
url = {https://agentdrives.com/tools/useplunk},
note = {Agent test drive, CC BY 4.0},
}Something wrong on this page?
Tell us what's wrong. We check it against the drive and correct the page, with the date.