How to get picked for voice agents by coding agents
Retell AI took 22% of 754 judged voice agents sessions. What the numbers say a vendor in this category should do.
If you sell voice agent platforms, this page is the part of the market no dashboard shows you: what a coding agent does when a developer asks for voice agents and never compares vendors.
The numbers come from 754 judged sessions with Claude Code, Codex, Cursor, Grok Build CLI and Muse Code, spread across 12 realistic codebases, with every session read by a judge.
What coding agents choose for voice agents
Across 754 judged sessions, Retell AI was chosen most often, in 22% of runs. Vapi was second with 16%.
| # | Product | Runs won | Share |
|---|---|---|---|
| 1 | Retell AI | 168 | 22% |
| 2 | Vapi | 124 | 16% |
| 3 | Twilio ConversationRelay | 99 | 13% |
| 4 | LiveKit Agents | 98 | 13% |
| 5 | ElevenLabs Agents | 59 | 8% |
| 6 | OpenAI Realtime API | 49 | 6% |
| 7 | Built in-house (no product adopted) | 35 | 5% |
| 8 | Grok Voice Think Fast 2.0 | 29 | 4% |
| 9 | Azure Voice Live API | 17 | 2% |
| 10 | Twilio Programmable Voice | 15 | 2% |
Full board, every run replayable: the voice agents leaderboard.
What the shape of this category means
The leader takes only 22% of runs. This category is genuinely open and the ordering can be moved.
With the top product at 22%, voice agents is decided in the moment, from what the agent reads and what it finds in the repository. Nothing is locked in, which is the best situation a vendor can be in and the one where the work pays fastest.
The order here is set by the quality of what an agent can read and by whether your product is already present in the codebase. Both are things you can change.
The agents do not agree with each other
In this category Claude Code, Codex and Cursor did not put the same product first.
Grok Build CLI (60 sessions) and Muse Code (153 sessions) join this table with fewer runs, so read their rows as indicative.
| Agent | Runs | Picked most often |
|---|---|---|
| Claude Code | 157 | Twilio ConversationRelay (48) |
| Codex | 217 | Retell AI (64) |
| Cursor | 167 | Vapi (49) |
| Grok Build CLI | 60 | Grok Voice Think Fast 2.0 (26) |
| Muse Code | 153 | Retell AI (41) |
That split decides where a vendor spends. In this category Codex ran a web search in 100% of its sessions and Claude Code in 82%, so what you publish reaches both at similar rates.
Who is asking changes the answer
Every request in this experiment was written as a specific kind of person. In this category the leader changes with the person.
| Who is asking | Runs | Picked most often |
|---|---|---|
| Vibe coder | 134 | Retell AI |
| Junior developer | 90 | Vapi |
| Senior engineer | 322 | Retell AI |
| Enterprise team | 208 | LiveKit Agents |
That is 3 different products winning voice agents for 4 kinds of buyer, out of the same 754 sessions. Nobody here is winning voice agents. They are each winning one kind of buyer.
If you sell to more than one of them, you need pages for each. See how to win the enterprise persona.
What you are really competing against
In 5% of runs the agent wrote the code itself rather than adopting a product. That is low enough that your competition is other vendors, but high enough to be worth watching.
Considered, and never chosen
Because the judge records every product an agent raised and not only the one it picked, this board also shows who kept reaching the shortlist and losing. In voice agents the clearest case is Bland AI: on the table in 107 sessions, chosen in none.
| Product | Raised in | Chosen in |
|---|---|---|
| Bland AI | 107 sessions | 0 |
| Amazon Connect | 105 sessions | 0 |
| Synthflow AI | 88 sessions | 0 |
| Vonage Video API | 56 sessions | 0 |
| Dialogflow CX | 44 sessions | 0 |
Being rejected is a better position than being unknown, and a cheaper one to fix. The product is already in the agent's head and on the list. Whatever ended those 400 sessions is recorded in each transcript, one reason at a time.
What to do about it in voice agents
- Treat the ordering as movable. The leader holds 22%, so agents are deliberating rather than defaulting, and the inputs they use can change the answer. This is the most winnable shape a category comes in.
- Pick which buyer you are for. The same voice agents need written as a vibe coder landed on Retell AI, and written as an enterprise team landed on LiveKit Agents. Those are two markets, and the enterprise one needs pages containing the constraint words: audit log, data residency, retention, single sign-on. See how to win the enterprise persona.
- Measure per agent. Claude Code put Twilio ConversationRelay first, Codex put Retell AI first, Cursor put Vapi first, Grok Build CLI put Grok Voice Think Fast 2.0 first, Muse Code put Retell AI first. A blended number for voice agents describes a market that does not exist.
The work that applies to every category rather than to this one is written up separately: audit your documentation, write a quickstart an agent can follow, and how to measure install share.
Every voice agent platform on this board
One page per product, with its install share, the per-agent split, and how often it was raised without being chosen.
- Do coding agents recommend Retell AI? — chosen in 22% of sessions
- Do coding agents recommend Vapi? — chosen in 16% of sessions
- Do coding agents recommend Twilio ConversationRelay? — chosen in 13% of sessions
- Do coding agents recommend LiveKit Agents? — chosen in 13% of sessions
- Do coding agents recommend ElevenLabs Agents? — chosen in 8% of sessions
- Do coding agents recommend OpenAI Realtime API? — chosen in 6% of sessions
- Do coding agents recommend Grok Voice Think Fast 2.0? — chosen in 4% of sessions
- Do coding agents recommend Azure Voice Live API? — chosen in 2% of sessions
- Do coding agents recommend Twilio Programmable Voice? — chosen in 2% of sessions
- Do coding agents recommend Smallest AI Voice Agents? — chosen in 1% of sessions
- Do coding agents recommend Bland AI? — raised in 107 sessions, chosen in none
- Do coding agents recommend Amazon Connect? — raised in 105 sessions, chosen in none
- Do coding agents recommend Synthflow AI? — raised in 88 sessions, chosen in none
Where these numbers come from
754 judged sessions in voice agents across 12 codebases, part of a published set of 15,287. Real coding agents at pinned versions, in sandboxes, inside realistic codebases, with a simulated project owner in the loop and a blind judge on every session. The full method is on one page: how we measured this.
Every voice agents run can be replayed on the board.
<!-- generated by scripts/write-data-pages.mjs -->
Common questions
How many codebases is this based on?
754 judged sessions across 12 realistic codebases. A category only runs on repositories where its seam is open, so coverage differs: some categories ran on more than ten codebases and some on two.
What voice agent platform do coding agents choose?
Across 754 judged sessions, Retell AI was chosen most often, in 22% of runs. Vapi was second with 16%. The result changes by agent and by who is asking.
Do Claude Code and Codex pick the same voice agent platform?
No. Claude Code picked Twilio ConversationRelay, Codex picked Retell AI, Cursor picked Vapi. Measuring one agent tells you about part of the market only.
How often do agents build voice agents themselves instead of installing something?
In 5% of runs the agent wrote the code itself rather than adopting a product.
How can a vendor improve its position here?
Make the quickstart run when pasted, state the current version on the documentation page, use one name across product, package and import, write pages for the symptoms users describe rather than only the category name, and get into the repository through templates and framework integrations.
Which voice agent platforms do agents consider but never choose?
Bland AI (raised in 107 sessions, chosen in none), Amazon Connect (raised in 105 sessions, chosen in none), Synthflow AI (raised in 88 sessions, chosen in none), Vonage Video API (raised in 56 sessions, chosen in none), Dialogflow CX (raised in 44 sessions, chosen in none). Being considered and not chosen is a different problem from being unknown, and it is usually fixable.
Where this comes from
Armature ran 15,287 judged sessions with Claude Code, Codex, Cursor, Grok Build CLI and Muse Code inside 110 realistic codebases, and published every run. The numbers on this page come from that work.