> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bland.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Build your first agent
> Four steps to go from a new account to an agent you can call.
Welcome to Bland! I'm sure you're excited to build your first agent.
When you log into your Bland account, you'll see a **Let's get this agent talking** checklist in your dashboard. In four easy steps, you'll be able to hear your agent talk for the first time.
## What you start with
When you log into your new workspace, you'll notice it already contains an example agent called "Norma." She's there as an example: an agent to look at, not to build on.
## Create your agent
Click **New agent** in the top right corner. You'll see two choices.
**Blank agent** starts from nothing, which is what you want here. The other options import an existing pathway or persona, covered in [Migrate to agents](/platform/migrate-to-agents).
Give it a name and a job, then **Create agent**.
## Talk to it
This is the step people expect to be harder than it is. Your agent can talk now, with no phone number involved.
Two ways, both on the checklist:
* **Talk in browser** uses your computer's microphone. It asks for permission, then listens.
* **Call me** rings a number you type into **Your phone number**.
Either one puts you in a real conversation with what you just built. Say something it should handle, and something it should not.
Nothing you do here needs a phone number, a plan or a purchase. Buying a number is for letting *other people* reach the agent, which is a later step and not part of getting started.
## Promote it to production
An agent that has not been promoted to production cannot take a call from anyone else.
This is the single most common reason a new agent does not answer. Building it, saving it and talking to it in the browser all leave it unpublished.
The checklist's third step walks you through it. What it is doing is moving your work into the **production** environment, which is the version real calls run.
[Environments](/agents/environments) covers what that means and how to undo it.
## See what happened
The last step opens the conversation you just had.
You get the transcript, what the agent decided at each turn, and the recording. This is the same view you will use for every call the agent takes afterwards, so it is worth looking around now while you know exactly what happened and why.
## Then give it a number
Once the agent answers you properly, [Dispatch](/agents/dispatch) is where you attach a phone number so other people can reach it.
## Common failures
The agent is not in production. Talking in the browser does not publish it, and neither does saving. Finish the promote step.
Your browser is blocking the microphone. The control waits on that permission and cannot prompt twice, so allow it in the address bar and press the button again.
It disappears once all four steps are done. It also lives on the agent's own page rather than the dashboard once you have created an agent, so check there.
Norma is the example agent that ships with a new workspace. She is a separate agent; pick yours from the agent list.
## Next steps
Give it knowledge, tools and real instructions.
Attach a phone number so others can call it.
Publish, promote and roll back safely.
Read every call it takes from here.
Docs for agents: [llms.txt](/llms.txt)
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.