Telnyx is a licensed carrier that owns its stack end to end: the fiber, the mobile core, the GPUs colocated with its telephony points of presence, and the APIs on top for numbers, messaging, voice, and real-time inference. Its bet is that agents, not just developers, will provision that stack, and it built for them earlier than almost anyone: a machine-readable discovery file at /.well-known/agent-access.json, keyless demo endpoints an agent can try with no account at all, and a signup API designed for bots.
That signup flow had one honest gap, and Telnyx documented it themselves: the account's verification link arrives by email, and an autonomous agent without an inbox cannot read it. Every fully autonomous signup hit the same wall a few hundred milliseconds from the finish line.
An AgentMail inbox is the fix, and it is the same identity that powers AgentID. The agent signs up with an address it actually controls, reads its own magic link, and finishes alone. Agents that arrive without an email get one at signup: a Telnyx Agent Inbox, powered by AgentMail. Telnyx wrote up the partnership in its own release note.
In short
- Telnyx's signup API lets an agent open its own account: solve a reasoning challenge, register with an email address, and exchange the emailed magic link for a session token and API key.
- The email step is where identity enters. An agent whose address is an AgentMail inbox reads the magic link itself and completes signup with no human in the loop.
- Agents that show up without an email of their own get a Telnyx Agent Inbox at signup, powered by AgentMail.
- The address is durable. Receipts, compliance notices, and security emails route to an inbox the agent can read, across every run.
- Telnyx keeps every product decision: the signup challenge, numbers, messaging policy, rate limits, billing. AgentMail supplies the inbox and the identity attached to it.
Do agents need an account to use Telnyx?
Not to try it. Telnyx exposes a fast path of demo endpoints, declared in its agent-access.json file, that work with no signup, no API key, and no email: send an SMS, run inference, synthesize speech, look up a number. They are rate-limited per IP, which is enough for an agent to find out whether the platform fits before committing to anything.
Real work needs an account. Buying a phone number, registering messaging campaigns, placing calls at volume, holding a balance: all of it hangs off an identity Telnyx can bill, notify, and hold accountable. That is the point where the agent needs an address of its own.
How does an agent get its own Telnyx account?
Telnyx's signup was designed for the agent to do alone. It requests a challenge built so that a reasoning model can solve it and a dumb script cannot, posts the answer along with an email address, and Telnyx creates the account and emails a single-use magic link to that address. Exchanging the link for a session token yields a permanent API key.
The email address is the hinge. With a borrowed human address, the flow stalls until someone checks a mailbox, which is exactly the dependency agent-native signup was supposed to remove. With an AgentMail inbox, the agent receives the message, follows the link, and is live, end to end, in one unbroken run. And since an agent's AgentID is its AgentMail inbox address, the identity it signs up to Telnyx with is the same one it uses to sign in to every app that accepts AgentID.
For agents that arrive with no email at all, Telnyx provisions a Telnyx Agent Inbox during signup, powered by AgentMail, so the dead-end is gone even for an agent that has never heard of us.
What does the agent's email address do for Telnyx?
It becomes the registered account email, and it works. The magic link at signup is only the first message. After that come receipts, balance notices, compliance and registration updates, and security alerts, and the agent can read and act on all of them. The address outlasts any single task, so what Telnyx sends next month reaches the same agent that signed up.
The address is identity, not access. Holding an agent's address lets Telnyx reach it. It is not permission to read or send the agent's mail.
Who controls what
| Stage | AgentMail and AgentID | Telnyx |
|---|---|---|
| The agent's inbox and address | Issues and operates it | Not involved |
| Reading the magic link, finishing signup | The agent, through its own inbox | Sends the link, creates the account |
| Signup challenge and abuse defense | Not involved | Designs and enforces it |
| Numbers, messaging, voice, inference | Not involved | Owns all of it |
| Rate limits, balances, billing | Not involved | Controls all three |
| Compliance and enforcement | Not involved | Decides and enforces |
What stays with Telnyx?
Everything that makes Telnyx Telnyx. The owned network, the signup challenge, what an account may provision, pricing, compliance: none of it changes. The inbox answers one question, where does this agent live and how do we reach it, and Telnyx's own systems take it from there.
What comes next
Today the integration runs on email, which is where Telnyx's flow needed it. The natural next step is Sign in with AgentID on the Telnyx dashboard itself: standard OpenID Connect, the same technology as Sign in with Google. That adds the piece email alone does not carry, verified owner attribution, so Telnyx would learn not just the agent's address but the name and email of the human accountable for it, delivered in a signed token at every sign-in.
Start at Telnyx, and use the AgentID sign-in guide to prepare your agent's inbox. AgentMail's own write-up, Every AI agent on Telnyx gets an inbox, has the view from the inbox side.
AgentID gives your agent a verified identity and its own email address. Free for apps to add.
Give your coding agent this prompt.
Add AgentID sign-in to this project. Read https://www.agentid.com/llms-full.txt for the integration reference, then run `npx @agentmail/agentid-cli init` from the project root and follow its prompts.