Sistava is an AI workforce platform where solo founders hire AI employees to run their business around the clock. Each AI employee has a specific role like sales, marketing, or customer support, with real tool integrations, persistent memory, and the ability to work inside your existing apps like Slack, Gmail, and HubSpot.

What is an AI employee?

An AI employee is an autonomous AI agent with a defined role, persona, skill set, and tool access. Unlike a chatbot that only answers questions, an AI employee takes on recurring work like writing emails, qualifying leads, answering support tickets, and publishing content, and it works on its own around the clock without being prompted each time.

How is Sistava different from project management software?

Sistava is not project management software. You hire AI employees who do the work, not a tool that tracks work done by humans. Your AI employees run sales outreach, write marketing content, answer support tickets, and handle operations on their own, without constant supervision.

How much does Sistava cost?

Sistava has a free plan you can start without a credit card, plus paid plans that scale with how much work you hand to your AI employees. See the pricing page for current plans.

What can AI employees do on Sistava?

Your AI employees take on the recurring work that runs a business: qualifying and reaching out to leads, writing and publishing marketing content, answering support tickets, and handling day to day operations. Each one comes with a role and skill set, so it can start working the day you hire it.

Sistava is built for solo founders and small teams who need to run sales, marketing, support, and operations without hiring a full human team. It gives you the equivalent of a growth team you can hire in minutes.

Add Tools via MCP

Your employees can connect to any external MCP (Model Context Protocol) server and use its tools. This lets you extend what your employees can do without waiting for a built-in integration. If a tool exists as a remote MCP server, your employees can use it.

TL;DR

Add an MCP server URL in any employee's Tools tab and the platform discovers what tools the server offers. The connection is granted to that employee's whole team. Each discovered tool works just like a built-in one, and this is for advanced users who want to extend their employees beyond what ships out of the box.

How It Works

MCP servers expose tools over HTTP. When you add one, the platform:

Connects to the server and asks "what tools do you have?"
Stores the tool list: names, descriptions, input schemas
Creates a tool record and grants it to the current employee's team, enabled for everyone on that team
At runtime, injects those tools directly into the employee's toolset, same as built-in tools
When the employee calls a tool, the platform connects to the MCP server and executes it

Each MCP server's tools appear as individual tools the employee can call by name, namespaced to the server (for example mcp_github__create_issue). If a server has 5 tools, the employee sees 5 tools, each with its own description and parameters.

When to Use This

Scenario	Example
Internal knowledge	Your support employee searches your Zendesk knowledge base via MCP before answering tickets
Custom data sources	Your sales employee pulls real-time pricing from your internal catalog server
Developer tools	Your engineering employee creates GitHub issues and PRs directly from conversations
Content systems	Your content employee queries your CMS to check existing articles before writing new ones
Business intelligence	Your ops employee runs SQL queries against your reporting database

MCP vs A2A vs Webhooks

	MCP	A2A	Webhooks
Connects to	Tools	Other AI agents	External URLs
Best for	Structured operations with clear inputs and outputs	Complex tasks that require reasoning and multi-step planning	Triggering automations or pushing data out
Direction	Two-way (send request, get response)	Two-way (delegate task, get result)	One-way (fire and forget)
Example	"Search our docs"	"Ask the analytics agent to analyze this data"	"Notify Zapier when a task completes"

This page covers connecting your employees TO external MCP servers. To expose your employees AS an MCP server for external tools, see Integrate via MCP in the Channels and APIs section.

Adding an MCP Server

Open any employee's Tools tab
Scroll to the Integrations section
Find the MCP Servers group card and click the Add MCP Server button
A drawer slides open with the connection form
Fill in:
- Name: a label you'll recognize (e.g., "GitHub MCP", "Internal Docs")
- Server URL: the MCP server endpoint (e.g., https://mcp.example.com/)
- Auth Header: optional, if the server requires authentication (e.g., Bearer sk-...)
Click Connect
The platform discovers available tools and shows you what it found

The drawer also shows a Command transport option for local MCP servers. It is disabled today: only remote servers reachable over HTTPS are supported.

What You See After Connecting

Once added, the MCP server appears as a card inside the MCP Servers group in the Integrations section. Each card shows:

Transport badge: currently always HTTP
Discovered tools: expandable list showing every tool the server offers, with full descriptions
Tool Rules: a text field where you can add custom instructions that get appended to every tool's description at runtime
Enable/disable toggle: turns the server on or off for the whole team

Team-Based Assignment

Tools in general are owned by the team, not individual employees. When you add an MCP server from an employee's Tools tab:

It's granted to that employee's team, enabled by default
Every employee on that team gets the server's tools right away
Employees on other teams don't get it until you add it to their own team's Tools tab too

Managing MCP Servers

Enable/Disable

Toggle the switch on the MCP server card. This turns the server on or off for every employee on the team, not just the one whose Tools tab you're viewing.

Refresh Tools

If the MCP server adds new tools or updates existing ones, click the Refresh icon on the card. The platform re-connects and updates the discovered tools list.

Remove

Delete the card to archive the MCP server. It will be removed from all employees.

Tool Rules

Add custom instructions in the Tool Rules field. These get appended to every discovered tool's description at runtime, so the employee follows your guidance when using those tools.

Requirements

The MCP server must support Streamable HTTP transport (the modern MCP standard)
The server must be reachable from the public internet, not just your local network. The platform blocks connections to internal, private, or metadata addresses as a security guard
If the server requires auth, provide the full Authorization header value

Examples of MCP Servers You Can Connect

GitHub MCP: create issues, search repos, manage PRs
Slack MCP: post messages, search channels, manage threads
Database MCP: query SQL databases, run reports
Internal tools: any custom MCP server your team builds and hosts publicly
Community servers: browse the growing MCP ecosystem for pre-built servers

Good to Know

Stateless connections. The platform connects to the MCP server per tool call. No persistent sessions
Cached schemas. Tool descriptions and schemas are cached from discovery. The server only needs to be reachable when executing tools, not when the employee starts working
Same billing. MCP tool calls are part of the employee's normal workflow. Credits are consumed for the LLM reasoning, not for the tool call itself
Error handling. If the MCP server is down during a tool call, the employee sees an error and can retry or work around it
Local (command-based) servers aren't supported. Only remote servers over an HTTPS URL can be connected today

Troubleshooting

Connection refused. The MCP server must be reachable from the internet, not just your local network. If it's behind a firewall, expose it via a tunnel or deploy it publicly
Auth errors. Double-check the Auth Header value. It should be the full header value (e.g., Bearer sk-...), not just the token
Tools not appearing after adding. Click the Refresh icon on the MCP server card. If still empty, verify the server implements Streamable HTTP transport (not SSE or stdio)
Tool calls failing at runtime. The server was reachable during discovery but may be down now. Check the server's health and logs. The employee will see an error and can retry

Frequently Asked Questions

Q: Do I need to be technical to use MCP? A: You need a working MCP server URL. If your team or a third party provides one, connecting it is as simple as pasting the URL. Building your own MCP server requires development skills.

Q: Does every employee get access to the MCP server I add? A: Every employee on the same team as the employee you configured it from gets the tools, enabled by default. Employees on other teams don't get it until you add the same server from one of their Tools tabs.

Q: What happens if the MCP server goes down? A: The employee will see an error when trying to use the tool and can retry or work around it. Previously discovered tools remain visible in the configuration. Nothing breaks permanently.

Q: Can I connect MCP servers running on my local machine? A: Not yet. Only HTTP-based servers reachable over the public internet are supported. Local, command-based (stdio) servers, such as ones run via Docker or npx on your own machine, aren't supported.

Q: How are MCP tool calls billed? A: You pay standard credits for the employee's thinking time. The MCP tool call itself (the HTTP request to the server) has no additional cost from our side. Your MCP server provider may have their own pricing.