Connected Agents

Your agents, our infrastructure

Bring the AI agents you built or bought into Plain. They work alongside your team with full customer context, and you decide what each agent can see and do.

Your agents do the work. Leave the orchestration to us.

Use the agent you already trust

You aren’t limited to any agent. Bring the ones your team already bought or built and trust on your own harness, model, and prompts.

Connect your agents in minutes

Your team wants to build agents, not support infrastructure. Connect your agents to our API, and let them work on the platform thousands of B2B teams already rely on.

Keep sensitive actions inside your systems

Your agents run on your own stack, so actions like refunds remain secure and you control the risk, evals, and spend. See how each agent performs and switch anytime.

“Plain is the only reason we can run a third-party AI support agent at all. We tried a lot of other support tools, and none came close to this level of flexibility.”

“Plain is the only reason we can run a third-party AI support agent at all. We tried a lot of other support tools, and none came close to this level of flexibility.”

“Plain is the only reason we can run a third-party AI support agent at all. We tried a lot of other support tools, and none came close to this level of flexibility.”

Connect any type of agent.

Build your own agent on Claude, Codex, or any other model, or connect third-party agents like Sierra, Decagon, Parahelp, and Decimal. Most teams start with one of these.

Joseph Myers

How do I rotate my API Keys?

Customer Agent

AI agent

Go to Settings, API keys. I’ve linked the steps for your plan.

Customer agents

Answer customers faster, on every channel

Agents that reply to customers and handle threads end to end on every channel, so your team can focus on the most complex threads.

T-2088

API Issues

Investigating thread…

Checked logs, 3 errors

Linked issue in Linear

Drafted note for the team

Internal agents

Give your team the answer before they open the thread

Agents that investigate issues, work through tasks, and keep your knowledge docs accurate. They don’t message customers, so they access internal tools.

Specialists

Sales Agent

Triage Agent

Insights

Knowledge gaps

Trend reports

Escalations Agent

Specialist Agent

Your next agent

Triage

Triage

Trend reports Agent

Knowledge Agent

Theme Reports

Your next agent

Knowledge gaps

Reporting Agent

New Agent

Triage

Trend reports

General agents

Put agents on any support work you can define

Triage new threads, report on support trends, close knowledge gaps, or run several specialist agents together — all with data from Plain.

Want a starting point? Check out our documentation.

“Bringing Fin into Plain meant we could roll it out across Slack, MS Teams, and any future channel without starting over.”

“Bringing Fin into Plain meant we could roll it out across Slack, MS Teams, and any future channel without starting over.”

“Bringing Fin into Plain meant we could roll it out across Slack, MS Teams, and any future channel without starting over.”

How to build your first agent.

// Plain posts thread events to your endpoint

// Plain posts thread events to your endpoint

app.post(“/plain/webhook”, async (req) => {

app.post(“/plain/webhook”, async (req) => {

  const { thread } = req.body.payload;

  const { thread } = req.body.payload;

  // Your model, your tools, your stack

  // Your model, your tools, your stack

  const answer = await myAgent.run(thread);

  const answer = await myAgent.run(thread);

  await plain.replyToThread({

  await plain.replyToThread({

    threadId: thread.id,

    threadId: thread.id,

    textContent: answer,

    textContent: answer,

  });

  });

});

});

Start from working code

Copy an example agent and adapt it to your stack. It already handles webhooks from Plain and replies through our GraphQL API.

✻ Welcome to Claude Code
Build me a customer agent like Plain’s example agent, for our company, and deploy it to our infrastructure.
●Reading example repo
⎿Read(plain-example-agent)
●Creating machine user
⎿API key created
●Deploying
⎿Deployed to your infrastructure

Have an LLM build it for you

Give Claude our example agent and ask for one built for your company and deployed to your infrastructure. With a working example to follow, it can build the first version in one go.

“Plain gave us the API surface to build automation that actually works. They became our support infrastructure, not just a configurable product.”

“Plain gave us the API surface to build automation that actually works. They became our support infrastructure, not just a configurable product.”

“Plain gave us the API surface to build automation that actually works. They became our support infrastructure, not just a configurable product.”

Connect your agent in three steps.

01

Create a machine user for each agent

Set its type to Agent and give it a scoped API key.

02

Subscribe to webhooks

Pick the events each agent needs.

03

Call the API

Your agents reply, add notes, set status, and hand off through our GraphQL API.

API-first from day one.

We built Plain so anything you can do in the app, you can do through the API.

We’ll keep Plain open to whichever agents work best for your customers.

Questions
and answers.

Everything you need to know about Connected Agents.

What are Connected Agents in Plain?

Connected Agents are AI agents a team builds or buys and connects to Plain through its API. They can reply to customers, help support teams investigate issues, or run in the background on work like triage and reporting.

Can I bring my own AI agent to Plain?

Which models and frameworks work with Plain?

Which third-party agents connect to Plain?

Where do connected agents run, and is that secure?

How do I see how each Connected Agent is performing?

Can I use my agents alongside Plain’s Customer Agent and Sidekick Agent?

How can my agents hand a thread to a human?

Get your support running on Plain in minutes.

Get your support running on Plain in minutes.

Get your support running on Plain in minutes.