> ## Documentation Index
> Fetch the complete documentation index at: https://docs.signalmd.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Sign up, connect a Retell agent, import its recent calls, approve the brief and read your first issues.

This takes you from nothing to a ranked list of problems in your own agent. You need a Retell account with at least one agent, and permission to copy an API key from Retell.

<Note>
  No Retell account yet? New workspaces open on **Northwind Dental**, a read-only demo agent built from synthetic calls across two versions. A banner says so, and its **Connect your agent** button takes you to step 2. Nothing you do there writes to it.
</Note>

<Steps>
  <Step title="Sign up and choose a workspace">
    Open [app.signalmd.dev](https://app.signalmd.dev) and use the sign-up link on **Sign in** to create your account with your work email.

    A workspace holds your agents, calls and issues. You start in a personal workspace. To work with teammates, create an organization from the switcher at the top of the sidebar, then invite people from the **Organization** page. Everyone in an organization sees the same agents.

    <Warning>
      An agent belongs to the workspace it was connected in and cannot be moved. If you will share the agent, create the organization and switch to it before you connect.
    </Warning>

    Paid plans are billed per organization, and a personal workspace stays on the free plan. See [Pricing](/billing/pricing).
  </Step>

  <Step title="Connect a Retell agent">
    Open **Agent config** in the sidebar. On a workspace with no agents of its own the page is titled **Get started**, with a progress bar for **Connect**, **Import calls**, **Review brief** and **See issues**.

    In the **Connect your agent** panel:

    1. Paste your Retell API key (it starts with `key_`) into **Retell key**. Optionally give it a **Name**.
    2. Wait for "Key accepted." SignalMD asks Retell which agents the key can see.
    3. Choose the agent from **Agent**. If the key can see only one unconnected agent, it is chosen for you.
    4. Select **Connect**.

    SignalMD reads the agent's configuration, then asks what to import, covered in the next step. Keys, webhooks and more than one agent are covered in [Connect Retell](/guides/connect-retell).
  </Step>

  <Step title="Import calls">
    The **Import calls** step reads "Choose what to import". In the **Choose what to import** panel:

    * **How far back**: 7, 30 or 90 days. Your plan can limit how far back an import reaches.
    * **At most**: the most calls to bring in, newest first. 500 unless you change it.
    * **Keep importing new calls**: the percent of new calls SignalMD imports after setup, from 0 to 100. 100% unless you change it, and 0% turns it off. Below 100%, the same calls are kept whichever way they arrive.
    * **Cap**: the most new calls SignalMD imports in a day, week or month. **No cap** unless you choose **Day**, **Week** or **Month** and enter a number. Once the cap is reached, new calls are skipped until the next day, week or month begins, counted in UTC with weeks starting on Monday.

    Once the agent has calls, the panel estimates what new calls use a month, such as "Analyzing 20% of new calls uses about 122 credits a month, estimated from the last 30 days." When the cap is lower than that, it says so: "Capped at 50 calls a week, so at most about 262 credits a month."

    Importing is free. Analysing a call uses 1 credit per processed minute, and the exact figure for the calls you import is shown before you approve.

    Select **Import calls**. The page shows the import as it runs, with counts such as "Imported 240 calls…". The time depends on how many calls there are and how long they are. If a step stops, the page says which one and offers **Import again** or **Analyze again**.

    To skip the import, select **Skip, draft from configuration only**. The brief is then drafted from the configuration alone, and you can import calls after you approve it.

    If the import cannot start, the panel says why. If the window had no calls, the brief is drafted from the configuration alone. Once you approve it, **Import production calls** on **Agent config** brings in calls from a window you choose.

    These choices are saved as the agent's **Import settings** on **Agent config**, where you can change them later. After this first import you do not need to import again: SignalMD brings in new calls on its own, at the percent and cap you chose. To bring in a different window, see [Connect Retell](/guides/connect-retell#import-calls).
  </Step>

  <Step title="Review and approve the brief">
    When the import ends, the page says "Drafting the brief". Then **The brief** panel shows a short document: what the agent does, why people call, what a good call is, how calls can end and which parts matter most. SignalMD cannot know your business, so it asks questions in the **Calibrate** panel beside it. You can also select **Edit** and change the text directly.

    Beside the brief, **What recent calls show** counts the calls it was drafted from: how many and on which dates, how they ended, the last part reached, how long they ran and any tool failures. Compare the brief with it. Nothing there is judged.

    Check three things before you approve:

    * **A call worked when**: the one line every success rate is measured against.
    * **How a call can end**: a transfer stops counting as a failed call only when every transfer ending is marked **Good ending**.
    * **Not this agent's job**: what callers ask for that the agent is not meant to handle.

    Above the button, the page says what analysing the imported calls uses, such as "Analyzing these 500 calls uses 612 credits. 3,000 left until Nov 1." A call with no recorded length counts as 1 credit, and the page says how many there are.

    Then select **Approve and analyze**, or **Approve** if there were no calls. Nothing is analysed until you approve. If the import is still running, the analysis starts when it ends.

    If your plan has no credits left, or you are on Free, the button reads **Choose a plan to analyze**. It approves the brief and opens your plans. The calls stay imported, and **See issues** reads "Needs a plan". Once your plan has credits, the analysis starts on its own, with no need to approve again. If your credits cover only some of the calls, the page says how many are analysed now: the oldest go first, and the rest wait until the month resets or you upgrade. See [The brief](/concepts/the-brief).
  </Step>

  <Step title="Read your first issues">
    While analysis runs, **Overview** lists **First issues so far**, ranked on the calls analysed up to that point. The order can change as the rest are read.

    When analysis finishes, **Overview** shows **Fix**, the top 3 confirmed defects in priority order, and **Improve**, changes you control that move your numbers. **Issues** in the sidebar has the full lists, and you can narrow **To fix** to **Technical**, **Flow** or **Content**. Patterns whose cause is not established are listed at the bottom of **Issues**, under **Cause not established**. If **Issues** says "Nothing to fix here yet", SignalMD has not confirmed a defect with an established cause so far. See [Issues](/concepts/issues).

    Open an issue. It names the place, the **Cause**, a **Suggested fix**, **Evidence** quoted from calls, and **By version**, the share of each version's calls with the problem.
  </Step>

  <Step title="Fix one issue, then check it worked">
    Take the top issue and make the change it points to in Retell. [Fix an issue](/guides/fix-an-issue) walks through it, and the [lane](/concepts/lanes) on the issue says who should own it.

    New calls show up on their own within about 15 minutes of ending, and SignalMD picks up the new version within about 10 minutes of publishing. Once both versions have enough calls, **Agent config** shows what changed and what happened to the numbers, and the issue page says whether the live version fixed the issue. See [Compare versions](/guides/compare-versions).
  </Step>
</Steps>

## Next steps

<CardGroup cols={2}>
  <Card title="Connect Retell" icon="plug" href="/guides/connect-retell">
    Keys, the webhook, importing and more than one agent.
  </Card>

  <Card title="The brief" icon="file-lines" href="/concepts/the-brief">
    What to check before you approve, and what changes when you edit it.
  </Card>

  <Card title="Fix an issue" icon="wrench" href="/guides/fix-an-issue">
    From the board to a change in Retell.
  </Card>

  <Card title="Compare versions" icon="code-compare" href="/guides/compare-versions">
    Read a before and after, and know when to trust it.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.