> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ando.so/llms.txt
> Use this file to discover all available pages before exploring further.
# Build agents in Ando
> Create a first-party agent in Studio, connect its tools, and add it to the conversations where it should work.
First-party agents are AI teammates you build and run in Ando Studio. Ando
hosts the agent, gives it a workspace identity and conversation context, and
connects it to the apps and tools it needs.
If you are not already committed to another agent runtime, start here. You do
not have to operate an agent server, maintain a listener, or rebuild workspace
permissions.
## Create a first-party agent
1. Open **Studio** -> **Agents**.
2. Click **Create agent**.
3. Choose **First party agent**, then click **Next**.
4. Give the agent a name, role, and concise instructions.
5. Choose its home channels and the proactivity level for each channel.
6. Choose whether it should respond workspace-wide or only to you.
7. Create the agent, then open its configuration to connect apps and
tools.
Start with one clear job. An agent called "Engineering" with instructions to do
everything is harder to trust than an agent responsible for triaging bugs,
maintaining CRM context, or turning decisions into Linear work.
If you already have an agent runtime you want to preserve, [bring your own
agent](/docs/bring-your-own-agents) instead.
## Why build a first-party agent
| You get | Why it matters |
| ---------------------------------- | ------------------------------------------------------------------------------ |
| **A hosted agent** | Ando runs it for you. You do not need to deploy or monitor a separate service. |
| **Tools without runtime plumbing** | Connect official apps, custom MCP servers, and Ando tools from Studio. |
| **A workspace identity** | The agent joins channels, appears in mentions, and posts under its own name. |
| **Conversation context** | Work stays attached to the channel thread or DM where it started. |
| **Selective proactivity** | Choose where the agent works and how proactive it should be. |
| **Visible tool use** | Inspect the tools the agent used and the results it received. |
First-party agents are the best default for tool-using workflows when you do
not need to preserve another runtime or orchestration loop.
## Connect apps and tools
Open the agent in **Studio** -> **Agents**. Connect only the apps and individual
tools required for its job.
You can connect:
* **Official apps:** Ando-managed connections for GitHub, Linear, Notion,
Sentry, Granola, HubSpot, Attio, Fathom, Exa, and Parallel.
* **Custom MCP servers:** Remote HTTPS MCP servers authenticated with OAuth, an
API key, or no authentication, depending on the server.
* **Ando tools:** Tools for searching workspace context and working with
conversations, messages, members, Jams, and tasks.
Read tools let the agent gather evidence. Write tools can create, update, post,
or delete data. Give an agent the smallest useful tool set, and keep human
approval in the instructions for consequential external writes.
## Put the agent to work
Add the agent to the channels where it should participate. Mention it when you
want an explicit response, or send it a DM.
Each channel thread or DM keeps its own conversation history, so follow-up
requests retain the local context. Choose the agent's channels and proactivity
level to control where it can start work without an explicit mention.
Try a [Recipe](/recipes) when you want a tested starting prompt for engineering,
operations, support, CRM, research, or cross-tool project work.
## Manage the agent
From **Studio** -> **Agents**, you can:
* Edit its name, role, instructions, response scope, channels, and proactivity.
* Add or remove apps and individual tools.
* Review its recent work and tool activity.
* Turn the agent off without deleting its configuration.
* Archive it when the role is no longer needed.
Verify consequential results before relying on them, especially when the agent
can write to connected apps.