Descubre folk el CRM para empresas impulsadas por personas.
Most AI tools can write a follow-up. Very few know which deal needs one, who owns the relationship, or what happened on the last call. folk CRM MCP closes that gap by connecting AI assistants directly to live contacts, companies, deals, notes, and interaction history inside folk.
The hosted Model Context Protocol server turns ChatGPT, Claude, Cursor, or VS Code into an operational AI Sales Assistant. Through secure OAuth access, sales teams can search relationship data, prepare account briefs, clean records, log conversations, create deals, and update pipeline stages from a natural-language prompt. The assistant does not work from pasted context or stale exports. It reads and writes within the CRM workflows the team already uses.
What Is folk CRM MCP?
⭐⭐⭐⭐⭐ (G2)
folk CRM MCP is a hosted Model Context Protocol server that gives compatible AI assistants secure read-and-write access to a folk workspace. It creates a standardized connection between the AI interface and live CRM records, removing the need for copied notes, CSV exports, or custom point-to-point integrations.
Once connected, an AI assistant can work with the same operational data as the sales team:
- People and companies
- Deals and pipeline stages
- Groups and custom fields
- Notes and interaction history
- Record ownership and relationship context
This access goes beyond conversational search. The assistant can retrieve an account’s history, create a new opportunity, change a deal value, move it to another stage, add a note, or clean incomplete records. Each request runs through the authenticated folk account and follows its existing workspace permissions.
A traditional CRM chatbot usually answers questions inside one application. folk CRM MCP takes the opposite route: it brings folk data and actions into tools such as ChatGPT, Claude, Cursor, and VS Code. The CRM remains the system of record while the chosen AI client becomes a practical command layer for daily sales work.
How folk CRM MCP Works?
folk CRM MCP sits between an AI client and the folk workspace. It translates a natural-language request into the relevant CRM search or action, then returns the result inside the same conversation.
AI client → folk MCP server → authenticated workspace → CRM result
1. One Hosted Server, No Custom Middleware
✅ folk hosts and maintains the MCP server. Teams do not need to deploy infrastructure, manage a local server, or build a separate connector for every AI platform. One connection gives supported clients a consistent way to interact with folk data.
2. Secure OAuth Authentication
✅ The connection uses OAuth rather than copied API credentials. Each user signs in to folk and authorizes the AI client to access the workspace. This keeps authentication tied to an individual account instead of exposing shared credentials in configuration files.
3. Access Based on Existing Permissions
✅ The AI assistant operates within the connected user’s workspace access. It does not create a separate database or bypass the CRM’s existing structure. People, companies, groups, notes, fields, and deals remain managed inside folk.
4. Read and Write Actions From One Conversation
✅ Read access lets the assistant search records, retrieve notes, review relationship history, or prepare a pre-call recap. Write access turns those findings into action: create a company, add a note, correct a field, open a deal, or move an opportunity to the next stage.
A sales rep could enter: Find open deals with no interaction in the last 30 days, summarize the latest notes, and flag the accounts that need a follow-up.
The assistant queries live CRM data and returns an actionable list without requiring the rep to filter several views manually.
What Can an AI Sales Assistant Do With folk MCP?
folk MCP connects natural-language instructions to concrete CRM actions. The most useful workflows combine retrieval and execution: the assistant finds the right context, interprets it, then updates the corresponding record.
folk CRM MCP: 5 Use Cases Across the Sales Cycle
A useful MCP workflow does more than execute an isolated command. It carries relationship context from one sales stage to the next, keeping research, conversations, and pipeline data connected inside folk.
1. Before Outreach: Build a Qualified Account List
Sales teams can describe a target segment in plain language instead of combining filters manually. folk MCP searches existing people and companies using location, role, industry, group membership, interaction history, or custom fields.
“Find SaaS founders in Germany with more than 20 employees, excluding anyone contacted during the last 60 days.”
The result becomes a focused prospect list based on CRM data, not a generic lead database. The same workflow can identify missing information before outreach begins.
2. Before a Call: Reconstruct the Relationship
A pre-call request can pull recent emails, meeting notes, deal data, stakeholders, objections, and promised next steps into one brief. This prevents reps from joining calls with fragmented context spread across several records.
“Summarize the relationship with Northstar, including the last three interactions, current deal status, open objections, and next agreed action.”
3. After a Call: Capture Details While They Are Fresh
folk MCP can turn a short written or voice recap into structured CRM updates. The assistant logs the conversation, records objections, updates relevant fields, and adds the next step to the correct contact or deal.
“Log the call with Maya at Northstar. Security approved the project, procurement needs the final quote, and the next meeting is on September 4.”
This reduces the lag between a customer conversation and an accurate pipeline.
4. During Pipeline Reviews: Find Revenue at Risk
Instead of reviewing every deal card, sales managers can ask for exceptions that require attention. Useful signals include stalled stages, missing next steps, declining engagement, overdue follow-ups, or deals without recent activity.
“List deals above $20,000 with no logged interaction in 21 days and group them by owner.”
The output creates a working review queue rather than another high-level pipeline summary.
5. After a Deal Moves: Keep Every Record Aligned
When an opportunity advances, folk MCP can update the stage, value, owner, notes, and associated records from one instruction. This prevents important changes from remaining buried in meeting notes or private messages.
“Move Northstar to Contract Review, set the value to $42,000, and add procurement approval as the next milestone.”
The CRM reflects the latest deal state immediately, giving the rest of the team reliable context for forecasting and follow-up.
How to Connect folk CRM MCP?
folk provides a hosted streamable HTTP endpoint:
https://mcp.folk.app/mcp
The first connection opens a secure OAuth flow. After authentication, the AI client receives the same folk workspace access as the connected account, based on its role and permissions.
Connect ChatGPT to folk
- Open Settings → Apps in ChatGPT.
- Open Advanced settings and enable Developer mode.
- Select Add Connector.
- Enter
https://mcp.folk.app/mcp. - Complete the folk OAuth authorization.
folk tools become available inside ChatGPT once the connection is active.
Connect Claude Desktop to folk
- Open Customize from the Claude sidebar.
- Browse the available connectors.
- Search for folk and add it.
- Sign in to the relevant folk workspace.
- Approve access through OAuth.
Claude can then search and update folk data from a conversation.
Connect Cursor to folk
Open Cursor Settings → MCP → Add new global MCP server, then add:
{
"mcpServers": {
"folk": {
"url": "https://mcp.folk.app/mcp"
}
}
}
Save the configuration and restart Cursor. The first folk action triggers the OAuth flow. Teams can also place the same configuration in .cursor/mcp.json at the project root.
Connect VS Code to folk
Create a .vscode/mcp.json file inside the workspace:
{
"servers": {
"folk": {
"type": "http",
"url": "https://mcp.folk.app/mcp"
}
}
}
Open the Command Palette, run MCP: List Servers, and start the folk server. Complete the OAuth flow when prompted.
Test the Connection With a Read Request
The first prompt should retrieve information without modifying a record:
“Find Acme Corp and summarize its latest notes, associated contacts, and active deals.”
Check that the assistant returns data from the correct workspace and respects the connected account’s permissions. Once the retrieval test succeeds, continue with a controlled write request:
“Add a test note to Acme Corp: MCP connection verified.”
Review the record directly in folk before allowing broader actions such as data cleanup, deal creation, or bulk field updates.
10 folk MCP Prompts Sales Teams Can Use Immediately!
Effective prompts specify the records to inspect, the business condition to apply, and the expected output or CRM action. Replace the example companies, stages, owners, and time periods with the fields used in the relevant folk workspace.
CRM Search
1. Find dormant relationships
“Find decision-makers in the SaaS group who have not received an email or logged an interaction in the last 90 days. Return their company, role, relationship owner, and last interaction date.”
2. Identify warm paths into an account
“Find every person connected to Northstar. Summarize the relationship history, identify the strongest existing connection, and explain why.”
Account Research
3. Prepare a pre-call brief
“Prepare a brief for tomorrow’s call with Acme. Include the stakeholders, latest interactions, active deals, objections, promised actions, and any notes mentioning security or procurement.”
4. Reconstruct a complex deal
“Summarize the full history of the Acme expansion deal in chronological order. Separate confirmed facts, open questions, buying signals, risks, and next steps.”
Data Hygiene
5. Audit incomplete company records
“Find companies in the Qualified group with a missing industry, website, location, or employee range. Return a table showing every incomplete field before making changes.”
6. Flag possible duplicates
“Find people who may be duplicates based on matching email addresses, names, companies, or LinkedIn profiles. List the suspected pairs for manual review without editing any records.”
Deal Management
7. Find stalled opportunities
“List open deals that have remained in the same stage for more than 21 days. Include deal value, owner, last interaction, current stage, and recorded next step.”
8. Update a deal after a call
“Move the Northstar deal to Contract Review, set its value to $42,000, add procurement approval as the next milestone, and log today’s call summary in the deal notes.”
Follow-Up Execution
9. Draft a context-aware follow-up
“Review the latest interactions with Maya at Northstar. Draft a concise follow-up that addresses the pricing concern, confirms the agreed security review, and proposes two next steps. Do not update the CRM yet.”
10. Build a prioritized follow-up queue
“Find open opportunities with no future next step. Rank them by deal value, relationship strength, and time since the last interaction. Explain the reason for each priority and suggest one follow-up action.”
For write actions, separating preparation from execution reduces accidental changes. A first prompt can request a preview, while a second confirms the exact records and fields to update.
Conclusión
folk CRM MCP turns live relationship data into an operational layer for AI. Through one hosted server, ChatGPT, Claude, Cursor, and VS Code can search records, prepare account briefs, update deals, clean data, and log sales activity directly in folk. The CRM keeps the context; the AI assistant helps the team act on it faster.
Preguntas frecuentes
Does folk CRM have a native MCP server?
Yes. folk provides a hosted MCP server that connects compatible AI clients directly to live CRM data through the Model Context Protocol.
Which AI assistants work with folk CRM MCP?
folk CRM MCP works with ChatGPT, Claude, Cursor, VS Code, Windsurf, Codex, and other clients that support remote MCP servers.
What can an AI assistant do in folk through MCP?
It can search contacts and companies, retrieve relationship history, create and update records, add notes, clean data, and move deals through a pipeline.
Is coding required to connect folk CRM MCP?
No coding is required for supported clients such as ChatGPT and Claude Desktop. Developer tools may require a short configuration containing the official folk MCP endpoint.
Is folk CRM MCP secure?
folk MCP uses OAuth and applies the permissions of the connected folk account. Teams should use trusted AI clients and require human approval for bulk or sensitive CRM changes.
Descubre folk
Como el asistente de ventas que tu equipo nunca tuvo