What is the Simmie MCP server?

It is a Model Context Protocol server with six tools: list_personalities, list_team_members, create_simulation, setup_scoring_rubric, score_transcript, and recommend_training. An MCP-compatible assistant can look up existing organization personas and people, create simulations, set the default rubric, score real call transcripts, and build targeted practice from those scores.

How do I connect Simmie to Claude, ChatGPT, or Copilot?

Sign in to your Simmie workspace and open the Connect MCP panel at https://www.simmie.co/#connect-mcp. Simmie generates a workspace-scoped server URL and access token. Paste both into your MCP client's server settings and the tools appear automatically. Step-by-step instructions for each client live in the Simmie support center.

Is there a cost to using MCP access?

MCP access is included on Teams and Enterprise plans. Sessions run through MCP consume the same practice credits as sessions started in the Simmie app, so there is no separate integration fee.

What is the Model Context Protocol?

MCP is an open standard that lets AI assistants and agents connect to external tools and data sources and invoke them directly. Because it is a standard rather than a bespoke integration, one Simmie MCP server works across every compatible client, and new clients can reach Simmie without additional engineering work.

Can our own internal agents call Simmie?

Yes. An MCP-compatible internal agent can call these tools with the connected account's permissions. For example, a team admin can confirm a rep and learner speaker label, score a supplied transcript with score_transcript, then pass the returned conversation_id to recommend_training. Your client should obtain approval before changing the organization-wide rubric or permanently saving a score.

What data does the MCP server expose?

list_personalities returns existing organization personas with ID, name, summary, profession, difficulty, and owner scope. It is read-only, capped at 50 results, and has no pagination. list_team_members is also read-only and returns people a scored call can be attributed to, including email addresses. Treat these results as organization data and share them only with authorized users.

Do reps need to install anything?

No separate MCP software is needed for reps who receive a practice link. A connected assistant calls create_simulation and returns a shareable link that the rep opens to practice. The current tool set creates simulations; it does not run a live roleplay inside the assistant.

How does this fit with the tools we already use?

Your MCP-compatible assistant or internal agent can supply a simulation brief or speaker-identified transcript and use Simmie's tools to create practice or score a call. These six tools do not retrieve transcripts or readiness records. Your existing system supplies the transcript, and recommend_training creates targeted practice from the resulting conversation_id, assigning it to the rep on team accounts.

Which tools require an admin on team accounts?

create_simulation, setup_scoring_rubric, and score_transcript are admin-only on team accounts. The two list tools are read-only. recommend_training builds practice from a scored call and assigns it to the rep on team accounts.

What changes are permanent or organization-wide?

setup_scoring_rubric silently replaces the organization's existing default scoring rubric, whether you adopt an existing rubric by title or create one from criteria. score_transcript permanently saves a scored call to a named person's record and cannot be undone. Supply turns with speaker and text, identify the learner's speaker label, and include for_user on team accounts. Build a confirmation step into your client before either operation.

How long do creation and training take, and can I retry?

create_simulation takes 45-60 seconds and always creates a new persona alongside the module and rubric, even if a similar persona already exists. It returns a shareable practice link. recommend_training takes up to two minutes and is safe to retry with the same conversation_id returned by score_transcript. Do not assume that retry guarantee applies to simulation creation or transcript scoring.

Live · Model Context Protocol

The Simmie MCP Server

Turn real calls into targeted practice from your AI assistant. Connect Simmie to Claude, ChatGPT, Microsoft Copilot, or your own agents to create shareable simulations, score transcripts against your organization's default rubric, and build training for the gaps.

Connect Simmie MCP See how it works

Connect at simmie.co/#connect-mcp

standard protocol: Openstandard protocol
callable tools: 6callable tools
read-only tools: 2read-only tools
Type II certified: SOC 2Type II certified

How It Works

MCP turns Simmie from a destination into a callable capability

The Model Context Protocol is an open standard that lets AI assistants and agents connect to external systems and invoke them directly. Think of it as a universal port: instead of building a one-off integration for every assistant, a platform exposes its capabilities once and any MCP-compatible client can call them.

Simmie exposes six tools: look up existing personas and team members, create simulations, set your default scoring rubric, score real call transcripts, and generate targeted practice from a scored call.

Client

Describe the practice you need

Ask your assistant to create a simulation for a renewal call with a skeptical CFO. Provide a plain-English brief and, optionally, an industry. On team accounts, an admin makes the request.

Discovery

The client reads Simmie's tool list

On connect, Simmie advertises its callable tools and the arguments each one accepts. The client decides which tool fits the request. No custom glue code.

Invocation

Create the simulation

The client calls create_simulation with your brief. Simmie builds a module, a new persona, and a rubric. Each call creates a new persona, even if your organization already has similar ones.

Return

Open the practice link

After 45-60 seconds, Simmie returns a shareable practice link to your assistant. Open the link to practice, or share it with your team. The roleplay itself does not run inside the MCP client.

Create practice from your assistant. Open the shareable link when it is time to rehearse.

Connect in Four Steps

Connect endpoint

Your workspace MCP endpoint and token are generated here:

https://www.simmie.co/#connect-mcp

Connect Simmie MCP

Need a hand with setup, tokens, or scopes? Our support center has the full walkthrough, or email support [@] simmie.ai.

Good to know

Included with Teams and Enterprise plans
Tokens are workspace-scoped and revocable any time
Nothing for reps to install

01
Open the connect screen
Sign in to your Simmie workspace and go to the Connect MCP panel.
02
Generate your server URL and token
Simmie issues a workspace-scoped MCP endpoint and access token. Both are shown once.
03
Paste into your MCP client
Add the endpoint and token to Claude, ChatGPT, Copilot Studio, or your own agent.
04
Call the tools
Your client discovers the six tools and their inputs. Ask it to create a simulation and return a shareable practice link. On team accounts, creation requires an admin.

Callable Tools

What your agent can actually do with Simmie

Six tools are currently available: two read-only lookups, simulation creation, default rubric setup, transcript scoring, and targeted training. Your client discovers each tool's input schema when it connects.

`list_personalities`

List the roleplay personas your organization already has, including ID, name, summary, profession, difficulty, and owner scope.

`list_team_members`

List people in your organization a scored call can be attributed to, including their email addresses.

`create_simulation`

Build a complete simulation with a module, persona, and rubric, and receive a shareable practice link.

`setup_scoring_rubric`

Set your organization's default rubric for scoring real call transcripts by adopting an existing rubric by title or creating one from criteria.

`score_transcript`

Score a speaker-identified transcript against the default rubric and permanently save it to a named person's record.

`recommend_training`

Build targeted practice for the weak areas in a scored call. On team accounts, the practice is assigned to the rep.

Example: from a real call to targeted practice

> "Score this call for Jordan using our default rubric, then create practice for the weak areas. Confirm Jordan's record and speaker label with me before saving."

Suggested sequence for a team admin

list_team_members identifies Jordan. Confirm the person and which speaker in the transcript is the learner.
Use the default scoring rubric. If one needs to be set, get approval before calling setup_scoring_rubric, since it replaces the organization-wide default without a warning.
score_transcript receives the speaker-labeled turns, learner label, and for_user. Once confirmed and called, the score is saved permanently.
recommend_training uses the returned conversation_id to create targeted practice and assign it to Jordan.

Allow up to two minutes for targeted training. If you need to retry recommend_training, use the same conversation_id rather than scoring the transcript again. Build confirmation steps into your client; the rubric replacement is silent and transcript scoring cannot be undone.

Compatibility & Governance

Where Simmie is callable

MCP is a standard, so Simmie works with whatever agent your organization already runs. No rip-and-replace.

Microsoft Copilot StudioLive
ClaudeLive
ChatGPTLive
Microsoft TeamsLive
Cursor & IDE agentsSupported
Custom in-house agentsSupported
CRM workflows (Salesforce, HubSpot, Dynamics)Supported
Any MCP-compatible clientSupported

Built for regulated teams

Opening a protocol does not mean opening your data. Access is scoped, logged, and revocable.

Workspace-scoped tokens

Every connection uses a token bound to your workspace. Revoke it any time without touching other integrations.

SOC 2 Type II

The same certified infrastructure and encryption that backs the core Simmie platform.

Full audit trail

Every tool call is logged with the caller, arguments, and result, so compliance can reconstruct any session.

Private deployment

Enterprise plans can run the MCP server against an Azure-hosted private LLM deployment.

MCP Questions, Answered

Looking for setup detail, troubleshooting, or client-specific steps? Visit the Simmie support center.

Make coaching callable from the tools your team already uses

Connect in about five minutes. If you would rather have us walk your team through it, book a time and we will set it up with you.

Connect Simmie MCP Talk to us

Setup guides and troubleshooting live in the support center. You can also reach us at support [@] simmie.ai.