Seismic MCP Server
The Seismic MCP Server exposes Seismic sales enablement capabilities as tools that AI agents and assistants can invoke natively. It implements the Model Context Protocol (MCP) specification version 2025-06-18 over Streamable HTTP transport — an open standard for connecting AI models to external data and actions — allowing agents to search content, generate LiveSend links, retrieve meeting intelligence, and look up CRM context directly within an AI conversation, without custom API integrations.
Early Access
The Seismic MCP Server is currently in Early Access. To get access, reach out to your Customer Success Manager.
Before You Begin
Ensure you have the following before connecting:
- A Seismic tenant with MCP enabled (contact your CSM if not yet enabled)
- An MCP-type app created in the Seismic App Registry Portal — this provides the OAuth client credentials needed to obtain a user access token
- A user access token with the
seismic.mcpscope (see Authentication) - Tool access enabled for your app by a tenant administrator (see Tool Permissions)
- An MCP-compatible client (see Supported Clients)
Supported Clients
The Seismic MCP Server has been validated with the following AI clients:
| Client | Notes |
|---|---|
| Claude (Anthropic) | Fully supported via Streamable HTTP transport |
| ChatGPT (OpenAI) | Fully supported via Streamable HTTP transport |
| Copilot Studio (Microsoft) | Fully supported via Streamable HTTP transport |
| MCP Inspector | Recommended tool for testing and debugging connections |
Any client that supports the MCP 2025-06-18 specification over Streamable HTTP transport will work with this server.
Endpoints
Tenant endpoint
https://mcp.seismic.com/v1/tenants/<tenant_name>
- Automatically resolves the tenant from the URL path
- Recommended for most integrations
Global endpoint
https://mcp.seismic.com/v1
- Uses central authentication
- The user is prompted to provide their tenant name during the OAuth flow
Quick Start
Add the following to your MCP client configuration, replacing <tenant_name> and <access_token>:
Claude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"seismic": {
"type": "http",
"url": "https://mcp.seismic.com/v1/tenants/<tenant_name>",
"headers": {
"Authorization": "Bearer <access_token>"
}
}
}
}
Generic MCP HTTP client
POST https://mcp.seismic.com/v1/tenants/<tenant_name>
Content-Type: application/json
Authorization: Bearer <access_token>
The server responds using the standard MCP Streamable HTTP protocol. See MCP client documentation for client-specific setup details.
Authentication
All requests require a Seismic JWT user access token.
Authorization: Bearer <access_token>
Scope requirements:
| Scope | Purpose |
|---|---|
seismic.mcp | Required to establish a connection with the MCP Server |
The
seismic.mcpscope alone does not grant permission to invoke tools. Tool invocation requires additional permissions configured by your tenant administrator. See Tool Permissions.
Obtaining OAuth Credentials
- Create an MCP-type app in the Seismic App Registry Portal
- Use the provided
client_idandclient_secretto request a user access token via the standard OAuth 2.0 authorization code flow - Include
seismic.mcpin the requested scopes
See Creating and Installing an MCP app in Seismic for step-by-step instructions.
Tool Permissions
The Seismic MCP Server uses a two-layer permission model:
Layer 1 — Tenant-level enablement
A tenant administrator must explicitly enable each tool for your MCP app in the Seismic App Registry Portal. Tools that are not enabled will return a TOOL_NOT_ALLOWED error when invoked.
Layer 2 — Feature availability
Some tools depend on Seismic features being active on your tenant (for example, Generative Search requires the Generative Search feature to be provisioned). If a required feature is not enabled, the tool will be unavailable regardless of permissions.
Administrator actions
Tenant administrators can manage tool access per MCP client app via the Seismic App Registry Portal under MCP App Permissions.
Available Tools
Search
Tool | Description |
|---|---|
| Get contents by Generative Search | Retrieves relevant sources for a user query without generating an answer. It Search and return the most relevant content recommendations ranked by relevance score |
| Generate answers or summaries by Generative Search | Generates a natural-language answer to a user question and returns the supporting sources used. |
Livesend
Tool | Description |
|---|---|
| Generate LiveSend link | Generates a LiveSend link with configurable settings and provided contents, enabling controlled sharing and engagement tracking. |
Meetings
Tool | Description |
|---|---|
| Get the meeting engagement list | Retrieve a paginated list of meeting engagements that match CRM context and filter criteria for a specific user, including basic metadata, ownership, and status suitable for search results and list views. |
| Get the transcript analysis of meeting | Get the transcript analysis of meeting by meeting id. In the text which was said by the speaker, A set of cue indicate that what the keywords and fuzzy matched keywords are against the what topics are during timeline. |
| Get post meeting overview | Get the post meeting overview, which include the summaries, the action items, content recommendations, objections, commercial topics and discovery questions from the AIML. Use this endpoint to quickly understand meeting results and prioritize follow-up without first retrieving full transcripts or detailed event payloads. |
| Get meeting list | Get a paginated list of meetings based on created time. Supports filtering by meeting type and pagination using offset, limit, and sortAscending. Use this endpoint to retrieve meetings for display in lists, search results, or analysis workflows. |
| Get meeting details by ID | Retrieves complete meeting information by meeting ID, including title, host, participant list, content presented, meeting status, and invitation context. This endpoint returns all essential metadata about a meeting that AI agents require for understanding meeting context, identifying participants, retrieving associated content, and accessing CRM integration data. The response includes invitees, actual participants, content items in presentation order, and meeting status (draft, scheduled, completed, etc.). |
DSR
Tool | Description |
|---|---|
| Get the DSR engagement list | Retrieve a paginated list of DSR engagements that match CRM context and filter criteria for a specific user, including basic metadata, ownership, and status suitable for search results and list views. |
CRM
Tool | Description |
|---|---|
| Get CRM context by ID | Retrieves CRM context information by context ID, including context name, type, system information, and associated metadata. This endpoint returns essential context details for understanding CRM objects like accounts, contacts, opportunities, and other CRM entities. The response includes context identification, display name, object type, and system identifiers used for cross-system integration and data correlation. |
| Get available CRM Systems | Retrieves a list of all available CRM systems integrated with the platform. This endpoint returns system metadata including unique identifiers, system types, and display names for each connected CRM instance. Use this endpoint to discover available CRM systems for context search operations, system-specific filtering, or multi-tenant CRM integrations. The response provides essential system information needed for cross-system operations and CRM context disambiguation. |
| Search CRM Contexts | Searches for CRM context objects using natural language queries and optional filters. This endpoint returns a list of matching CRM entities (accounts, contacts, opportunities, etc.) with confidence scores and match reasons. The search supports type hints to narrow results to specific object types and system hints to target specific CRM instances. Results include disambiguation indicators to help AI agents determine if additional clarification is needed from users. |
Tool Schema
Check the individual tool documentation for input and output schema information.
Error Reference
All errors follow a structured format with a code field identifying the specific failure.
{
"error": {
"code": "TOOL_NOT_ALLOWED",
"message": "The requested tool is not available for this tenant. Please contact your administrator to enable it.",
"correlationId": "a1b2c3d4-e5f6-7890-abcd-ef1234567890"
}
}
Reporting errors to support
When contacting support about an error, always provide the full response payload including the
correlationId. This is the primary identifier used to trace the request in server logs and is essential for diagnosis.
| Code | Category | Retryable | Description |
|---|---|---|---|
TOOL_NOT_ALLOWED | Validation | No | The tool is not enabled for this tenant or app. Contact your administrator. |
INVALID_TOOL_INPUT | Validation | No | A required input field is missing or invalid. |
SCOPE_NOT_ALLOWED | Auth | No | The access token does not include the required OAuth scope. Re-authenticate with the correct scopes. |
USER_NOT_ALLOWED | Auth | No | A user-level token is required. App-only (client credentials) tokens are not supported. |
INVALID_TOKEN | Auth | Yes | The token is invalid or expired. Obtain a fresh token and retry. |
TOKEN_EXCHANGE_FAILED | Auth | Yes | Internal token exchange failed. Re-authenticate and retry. |
DOWNSTREAM_TIMEOUT | Downstream | Yes | A Seismic service did not respond in time. Retry after a short wait. |
DOWNSTREAM_ERROR | Downstream | No | A Seismic service returned an error. Check the message for details. |
DEPENDENCY_UNAVAILABLE | Downstream | Yes | A required internal dependency is temporarily unavailable. Retry later. |
INTERNAL_SERVER_ERROR | System | Yes | An unexpected server error occurred. Retry, or contact support with the correlationId. |
Usage Examples
Content discovery and answers
Suggest some best practices for enterprise sellers, include sources
How is the SEP market evolving? Answer from my Seismic contents
Find the latest competitive battlecard for Salesforce
Content sharing with LiveSend
Generate a LiveSend link for the Q1 product overview deck
Create a shareable link for the Mars Case Study with tracking enabled
Meeting intelligence
Summarise my last meeting with Acme Corp
What objections came up in the Acme deal review last Tuesday?
Show me action items from yesterday's QBR
What topics were discussed in the meeting with ID abc-123?
CRM-connected workflows
What DSRs are associated with the Acme opportunity?
Show me the content shared in recent meetings for the Salesforce account
Find CRM context for opportunity ID 00646000009eMGL
Data & Privacy
- All tool results are scoped to content the authenticated user is authorized to access within Seismic. The server does not expose content outside the user's permissions.
- The MCP Server acts as a proxy to Seismic's existing APIs. No user data is stored by the MCP Server beyond what is logged for audit and observability purposes.
- Tenant isolation is enforced at the authentication layer — users can only access data within their own Seismic tenant.
- For full data processing and privacy details, see the Seismic Trust Center.
Support
For integration help, contact your Customer Success Manager or reach out via https://www.seismic.com/contact/.
Updated about 2 hours ago