Agent Personas

Instead of connecting your AI assistant to five different MCP servers one at a time, an Agent Persona lets you bundle the tools you need from across all your MCP servers into a single endpoint. Your team connects once, and gets access to everything.

Quick Start: What do you want to do?

I want to...	Jump to
Understand what an Agent Persona is and why I'd use one	What is an Agent Persona?
Create my first Agent Persona	Creating an Agent Persona
Connect my AI client (Cursor, Claude Desktop, etc.) to a persona	Connecting Your AI Client
Let a script or CI/CD pipeline use a persona	Generating an Access Key
Manage which apps are connected and authorized	Connecting Your Apps

---

What Is an Agent Persona?

Think of it this way: your organization has MCP servers for Slack, Jira, GitLab, Confluence, and more. Without Agent Personas, you'd need to add each one individually to your AI client, manage separate connections, and authenticate to each.

An Agent Persona wraps selected tools from any combination of these MCP servers into one endpoint. You connect your AI client to that single endpoint, authenticate once, and you're ready to work across all of them.

What it does for you:

• One connection instead of many. Your AI client connects to a single MCP endpoint, and the persona routes each tool call to the right MCP server behind the scenes.
• Pick only the tools you need. You don't have to include every tool from every MCP server — select just the ones relevant to your workflow.
• Everyone acts as themselves. When you send a Slack message or create a Jira ticket through the persona, it shows up as you — not a bot account.
• Team access control. Assign personas to specific teams so only the right people can use them.

---

Use Cases

Personal Productivity Agent

Create a persona for your day-to-day work that pulls together tools from the services you actually use.

Example: "Developer Workbench"

• GitLab — view merge requests, check pipeline status, browse code
• Jira — search issues, create tickets, update sprint items
• Slack — search conversations, post updates
• Confluence — find documentation, create pages

With this persona, you can ask your AI assistant things like:

• "What merge requests are assigned to me?"
• "Create a Jira bug for the login issue we discussed in #engineering"
• "Find the onboarding docs in Confluence and summarize them"

Automated Workflows

Create a persona for scripts, CI/CD pipelines, or scheduled jobs that need to interact with multiple services without human involvement.

Example: "Release Automation Agent"

• GitLab — trigger pipelines, list merged MRs
• Jira — transition tickets to "Done" on successful deploy
• Slack — post release notifications to #releases
• Confluence — auto-update release notes

Use an access key so your pipeline can authenticate without a browser.

---

Creating an Agent Persona

1. Select Agent Personas in the sidebar.
2. Select Create Agent Persona.

A three-step wizard opens:

Step 1: Basic Information

Field	Required	Description
Persona Name	Yes	A descriptive name — for example, "SRE Agent" or "Marketing Assistant"
Description	No	What this persona is for (helps your team understand its purpose)
Teams	No	Restrict access to specific teams. Leave empty to make it available to everyone.

Select Next.

Step 2: Tool Selection

You'll see two options at the top:

• AI Recommendations — Describe your workflow in plain language (for example, "I need to triage support tickets from Slack and track them in Jira") and the system suggests relevant tools from your available MCP servers.
• Manual Selection — Browse your MCP servers, expand each one, and check the individual tools you want to include. Use the search bar to find specific tools quickly.

Select the tools you need, then select Next.

tip

Only include tools that are necessary for the persona's purpose. Fewer tools means faster responses and a simpler interface for your AI assistant.

Step 3: Inbound Authentication

Choose how users and systems will authenticate to this persona. You can enable multiple options simultaneously:

Option	Best for	How it works
Interactive Login	Team members using AI clients (Cursor, Claude Desktop, etc.)	Users see an SSO prompt when they connect. Actions run under their own identity.
Agent Access Keys	Scripts, CI/CD pipelines, automation	A bearer token is generated that can be used directly — no browser needed.
Passthrough	External systems managing their own auth	Credentials are forwarded as-is. Optionally validate incoming JWTs.

For most teams, Interactive Login is all you need. Add Agent Access Keys if you also want to use the persona from automated pipelines.

Select Create & Deploy. Your persona is now live and ready to connect.

---

Connecting Your AI Client

After creating a persona, open it from the Agent Personas list. The Overview tab shows a Connect to MCP section with:

• Your MCP Endpoint URL (copy it with the button next to it)
• Ready-made configuration instructions for each supported client

Expand the section for your AI client — Claude Desktop, Cursor IDE, Windsurf IDE, Claude Code, or VS Code — and you'll see two setup options:

Option	What it does
Quick Setup (NPX)	Run a single terminal command that auto-configures your client. Easiest option.
Manual Setup	Copy a JSON snippet into your client's configuration file. Use this if you prefer manual control.

Example: Connecting through Cursor IDE

1. Open the persona detail page.
2. Expand Cursor IDE under "Connect to MCP."
3. Choose Quick Setup (NPX) and run the command shown in your terminal.
4. Restart Cursor. The persona's tools appear in your MCP tool list.

For Interactive Login, you'll be prompted to log in through your browser the first time you connect. After that, your session persists.

For Agent Access Keys, use the Manual Setup option and paste the bearer token into the Authorization header.

---

Generating an Access Key

Access keys let automated systems use the persona without a browser login.

1. Open the persona from the Agent Personas list.
2. On the Overview tab, find the Agent Access Keys section.
3. Select + Generate Key.
4. Give the key a Name (for example, "CI/CD pipeline - production").
5. Choose an Expiration: Never, 30 days, 90 days, or 1 year.
6. Select Generate.
7. Copy the bearer token immediately — it won't be shown again.

Expiration	When to use
Never	Long-running services (rotate manually on a schedule)
30 days	Testing, temporary integrations
90 days	Project-based automation
1 year	Production services

Using the key in your client

Add the token to your MCP client configuration:

{
  "mcpServers": {
    "my-persona": {
      "url": "https://your-persona-endpoint.a.run.app/mcp",
      "headers": {
        "Authorization": "Bearer your-access-key-token-here"
      }
    }
  }
}

Revoking a key

Go to the Authentication tab, find the key under Agent Access Keys, select the menu next to it, and select Revoke. The key stops working immediately.

---

Connecting Your Apps

Some MCP servers behind the persona use OAuth to access upstream apps (like Slack, GitLab, or Google Workspace). Before you can use those tools, you need to authorize each app — just like you would when connecting a new app on your phone.

Checking connection status

Open the persona and look at the Overview tab. The App Authentication section lists each MCP server and its status:

• Connect button visible — you haven't authorized this app yet. Select it to complete the OAuth flow.
• No button — the app is already connected and ready.

You can also check the Authentication tab, which shows the Connected Apps section with all MCP servers and their tool counts.

Completing the OAuth flow

1. Select Connect next to the app.
2. A browser window opens with the app's OAuth consent screen (for example, "Authorize Cequence to access your Slack workspace").
3. Approve the permissions.
4. You're redirected back. The status updates to connected.

This is a one-time step per app. After you authorize, the persona uses your stored credentials automatically for every future request.

Re-authenticating

If a connection expires or is revoked, select Connect again. The new credentials replace the old ones.

---

Managing a Persona

Viewing and editing

From the Agent Personas list, select any persona to see its detail page with three tabs:

Tab	What it shows
Overview	Details (name, status, teams, creator), App Authentication status, Access Keys, Connect to MCP instructions, and Available Tools list
Tools	All tools grouped by MCP server, each with an "Enabled" badge. Search to find specific tools.
Authentication	Agent authentication modes (Interactive, Access Keys, Passthrough) and Connected Apps

To change the persona's name, description, tools, or authentication settings, select Edit at the top of the detail page. The same three-step wizard opens with your current settings pre-filled.

To remove a persona entirely, select Delete.

---

Real-World Examples

Example 1: "I want a DevOps assistant for my engineering team"

Setting	Value
Name	DevOps Assistant
Tools	GitLab (MR management, pipelines), Jira (issue tracking), Slack (notifications), Confluence (docs)
Authentication	Interactive Login
Teams	Engineering

Each engineer connects from Cursor or Claude Desktop, logs in through SSO, authorizes GitLab/Jira/Slack/Confluence once, and then asks their AI assistant things like "What pipelines failed today?" or "Create a bug ticket for the auth regression."

Example 2: "Our CI/CD pipeline needs to post release notes automatically"

Setting	Value
Name	Release Bot
Tools	GitLab (list merged MRs), Confluence (create/update pages), Slack (post messages)
Authentication	Agent Access Keys
Teams	Platform

Generate an access key, add the bearer token to the pipeline's MCP client config, and the pipeline can compile merged MRs into a Confluence page and post a summary to Slack on every release.

Example 3: "Security wants a unified view of vulnerabilities across tools"

Setting	Value
Name	Security Analyst
Tools	Snyk (vulnerability scanning), Jira (issue tracking), Slack (alerting)
Authentication	Interactive Login + Agent Access Keys
Teams	Security

Security engineers use it interactively from their IDE. The same persona is also used by a scheduled script that runs nightly scans and posts high-severity findings to #security-alerts — using an access key.

---

Tips

• Start with one use case. Build a persona for your most common workflow first. You can always add more tools or create additional personas later.
• Use AI Recommendations. When creating a persona, describe what you want to do in plain language — the system suggests the right tools, saving you from browsing through every MCP server manually.
• Keep tool counts reasonable. A persona with 10–15 focused tools performs better than one with 100+ tools from every MCP server in your organization.
• Set up both auth modes when needed. Enable Interactive Login for your team and Agent Access Keys for automation — both can be active at the same time.
• Rotate access keys. For production pipelines, set an expiration and rotate keys on a schedule. You can generate a new key before revoking the old one to avoid downtime.
• Use team assignments. If a persona contains sensitive tools (for example, HR data, financials), assign it to specific teams instead of leaving it open to everyone.