Deepgram
deepgram.comDeepgram sells speech-to-text, text-to-speech and voice agents.
Can an AI agent use Deepgram? Yes, without signing up. Our agent got into Deepgram on its own in 20 minutes and did 5 of its 9 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 and an OpenAPI.
373 doc pages read
- On its own
Get in
Got in without signing up. Email and password: stopped by a CAPTCHA.
20 minutes
- Partly
Use
Did: list every speech and voice model over the API without a key; transcribe sample calls and the agent's own streamed audio in the free playground. Not done: transcribe own audio files or call the API.
5 of 9 core tasks
- Not reached
Pay
Paying needs an account.
- On its own
Leave
Nothing to cancel, made nothing to delete.
Getting in
1 of 8 ways in worked without a person.
The way that worked
No key needed20 minutes
- An inboxNot needed
- A browserNeeded
- A personNot needed
- Agent on a serverAPI, MCP, CLINoStops at the first API call: it needs a browser.
- Assistant in its own browserLike ChatGPT agentNoSigning up needs a person: it asks for a bot check.
- Signed agentKnown to CloudflareYesThe sign-up page opens.
Every way in
| Way in | What happened | Why |
|---|---|---|
| No key needed | Worked | |
| Email and password | Stopped by a CAPTCHA | |
| CLI code login | Stopped | The consent page needs an account |
| Sign-in for apps (OAuth, MCP) | Stopped | The consent page needs an account |
| Runs locally, no account | Not tried | Needs a person's account |
| Continue with GitHub | Not tried | Not part of this drive |
| Continue with Google | Not tried | Not part of this drive |
| Continue with Microsoft | Not tried | Not part of this drive |
What an agent can do with Deepgram
5 of 9 core tasks done, and 12 of 19 others.
Core tasks
- done there
- tried, not done
- offered, not tried
| Task | Result | Through |
|---|---|---|
| Transcribe streaming audio in real time | Done | CLI: offered, not tried |
| Transcribe a pre-recorded audio or video file | Done | CLI: offered, not tried |
| Convert text to speech in a single request | Done | CLI: offered, not tried |
| Run a real-time voice agent conversation | Done | Web app: done there |
| Detect speaker turns and interruptions during a call | Done | Web app: done there |
| Stream text to speech over WebSocketThe WebSocket needs a key. | Blocked | API: tried, not done |
| Dictate text into any Mac app | Blocked | Web app: tried, not done |
| Record a meeting without a bot joining the call | Blocked | Web app: tried, not done |
| Get a daily brief of calendar, email and messages | Blocked | Web app: tried, not done |
Reliability
76of 100
Core tasks that worked on all 3 runs
Handles mistakes
70of 100
Clear errors, retries without duplicates, honest limits
Surface tested
13of 100
Documented routes, tools and commands run and checked
Docs accuracy
84of 100
Documented operations that behaved as the docs say
Docs completeness
51of 100
How fully the docs describe each operation
What it can't do, and why
- Create an account The sign-up shows a visible CAPTCHA checkbox.
- Get an API key Keys come only from a Console account.
- Transcribe own audio files or call the API The API needs a key; file upload needs sign-in.
- Stream text to speech over WebSocket The WebSocket needs a key.
- Use Saga It is a Mac and iOS app with no web or API route.
- Pay Buying happens in the Console, behind the CAPTCHA sign-up.
12 of 19 other tasks done
- Diarize speakers in a transcriptDone
- Redact sensitive information from a transcriptDone
- Prompt custom vocabulary with keytermsDone
- Summarize a transcript or block of textDone
- Analyze sentiment of a transcript or textDone
- Detect topics in a transcript or textDone
- Recognize speaker intent in a transcript or textDone
- Detect named entities in audioDone
- Automatically detect or switch spoken language mid-conversationDone
- Choose the speech or voice modelDone
- Get a per-turn latency report from a voice agentDone
- List available speech and voice modelsDone
- Self-host Deepgram models in your own VPC or on-premPaid plan
- Bring your own LLM or TTS provider to a voice agentBlocked
- Register a callback URL for async transcription resultsBlocked
- Ask the meeting transcript a questionBlocked
- Learn a user's vocabulary and writing style over timeBlocked
- Create an API keyBlocked
- Invite a teammate to a projectBlocked
- Read usage and billing balanceBlocked
The plans they need
- Self-host Deepgram modelsEnterprise
Paid-plan tasks don't count toward the level. The drive used none (no free access: sign-up behind a visible reCAPTCHA); keyless playground only.
Surface audit: 165 documented operations, 28 run, 22 passed, 4 not as documented, 137 blocked (sandbox or plan limits).
How an agent connects
API, OpenAPI, CLI, SDK and llms.txt found.
MCP server
1 testedA server for searching the docs only.
3 tools across its docs
API
3 testedREST + WebSocket, 50 OpenAPI operations, 5 AsyncAPI channels.
55 operations across its docs
OpenAPI
Yes3.1.0; responses list only 200 and 400.
CLI
22 testeddeepctl 0.4.0 (dg), 23 commands, device-flow login.
52 commands across its docs
SDK
1 testedPython deepgram-sdk 7.16.0, JS @deepgram/sdk 5.17.0, plus Go, .NET, Java, Rust.
- PyPIdeepgram-sdk
- npm@deepgram/sdk
- PyPIdeepctl
48 methods across its docs
llms.txt
YesAgent payments
None foundThe checkout wasn't reached: Pay as you go.
Docs
No gaps in what we readCovers getting started, auth, error codes and code examples.
0 of 9 documented ways to do the core work worked when tried.
373 of 660 pages read
Tips for agent builders
- Read the docs as Markdown: append .md to any developers.deepgram.com page, or send Accept: text/markdown
- Use https://developers.deepgram.com/openapi.json and asyncapi.json for exact request shapes
- GET https://api.deepgram.com/v1/models lists models without a key
- For CI, pass DEEPGRAM_API_KEY instead of
dg login;dg mcpexposes the API to MCP clients with that key - Try features first in playground.deepgram.com: it shows the exact query parameters (e.g. summarize=v2, keyterm, language=multi)
- Without a key, a browser agent can still use the playground's streaming and voice agent pages by giving the tab its own audio as the microphone
- Check a model name against GET /v1/models first: a streaming session with an unknown model just fails to connect
How Deepgram compares
For the team behind Deepgram
Our agent got through Use only in part, and didn't reach Pay. Prove Deepgram 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.
7 things to fix for agents
Agent blockers: 3 foundAgent surfaces: 27 issues foundSafety review: 2 issues found
- FindClear
- Get inClear
- UsePartly
- PayNot reached
- LeaveClear
On our standard
27 criteria, each against every product we measured on it.
- Typed, tight input schemasAhead of most 50 of 65 others lower, 15 the same.
- Docs an agent can navigateAhead of most 98 of 141 others lower, 43 the same.
- Spend an agent can see and capAhead of most 85 of 118 others lower, 27 the same, 6 higher.
- The best way in is built for agentsAhead of most 98 of 147 others lower, 49 the same.
- One way to page through listsBehind most 53 of 87 others higher, 34 the same.
- Structured resultsBehind most 22 of 36 others higher, 14 the same.
- Export before leavingBehind most 89 of 144 others higher, 55 the same.
- Descriptions that say when to use the toolBehind most 46 of 65 others higher, 19 the same.
All 27 criteria
- Results sized for a context windowAhead of most 16 of 31 others lower, 15 the same.
- A tool list that fitsAhead of most 33 of 65 others lower, 32 the same.
- Truthful annotationsLevel with most 33 of 65 others the same, 32 lower.
- Sign-in for agents acting for a personLevel with most 64 of 119 others the same, 55 lower.
- Versions an agent can pinLevel with most 59 of 97 others the same, 38 lower.
- Answers sized for an agentLevel with most 61 of 85 others the same, 24 lower.
- Prices an agent can readLevel with most 80 of 105 others the same, 25 lower.
- Errors a program can readIn the middle 40 of 104 others lower, 43 the same, 21 higher.
- Fast to the first callLevel with most 71 of 85 others the same, 14 lower.
- Every way in works for an agentIn the middle 56 of 145 others lower, 51 the same, 38 higher.
- Tool names that sort themselvesLevel with most 61 of 65 others the same, 4 lower.
- A path written for agentsLevel with most 71 of 141 others the same, 39 lower, 31 higher.
- Real status codesLevel with most 125 of 127 others the same, 2 lower.
- Errors say what to do nextLevel with most 56 of 103 others the same, 24 lower, 23 higher.
- WebMCP tools on the web appLevel with most 91 of 93 others the same, 2 higher.
- A request id on every answerIn the middle 34 of 103 others lower, 29 the same, 40 higher.
- Agents pick the right toolLevel with most 44 of 65 others the same, 1 lower, 20 higher.
- All validation problems at onceLevel with most 23 of 35 others the same, 12 higher.
- Pages fit one fetchLevel with most 87 of 141 others the same, 54 higher.
Where Deepgram fits
- Wall
- Bot checksour agent got in another way
- Vendors
- Google reCAPTCHAon one way in
Found an error? Tell us on this page or read how corrections work.
Questions people ask about Deepgram
- Can an AI agent sign up for Deepgram on its own?
- Our agent got into Deepgram without signing up. Its way in: No key needed, in 20 min.
- Does Deepgram have an MCP server, and can an agent connect to it?
- Deepgram'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 1 Nov 2026, our agent got working access without signing up.
- How does an AI agent get a Deepgram API key?
- In the drive of 1 Nov 2026, our agent needed no key. Not needed: a documented route works without an account.
- Can an AI agent delete its Deepgram account?
- Not applicable: our agent used Deepgram without an account, and it made nothing to delete.
- Can an AI agent pay for Deepgram?
- No payment made: paying needs an account. No card needed to start.
Cite this page
The page as a reference, in plain text, APA and BibTeX.
Cite this
AgentDrives, “Deepgram: AI agent test drive”, 1 November 2026, https://agentdrives.com/tools/deepgram. CC BY 4.0.
AgentDrives. (2026, November 1). Deepgram: AI agent test drive. https://agentdrives.com/tools/deepgram
@misc{agentdrives_deepgram_2026,
author = {{AgentDrives}},
title = {Deepgram: AI agent test drive},
year = {2026},
month = nov,
url = {https://agentdrives.com/tools/deepgram},
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.