generated: '2026-09-13'
method: searched
source: >-
  https://github.com/PublicisSapient/knowhow-mcp (first-party source, Apache-2.0) and
  https://registry.npmjs.org/@psnext/lscg (first-party package README), read 2026-09-13
name: Publicis Groupe — Model Context Protocol servers
status: published
deployment:
  mode: local-stdio
  endpoint: null
  install: npm install -g @psnext/lscg && lscg mcp
  package: https://www.npmjs.com/package/@psnext/lscg
  auth: none
  verified: searched
  note: >-
    Publicis Groupe operates no vendor-hosted remote MCP endpoint. Both servers below are
    operator-installed: @psnext/lscg is a stdio server started with `lscg mcp` on the
    developer's own machine, and knowhow-mcp is a self-hosted Spring Boot container that
    an operator runs themselves — its transport is HTTP/SSE rather than stdio, but like a
    stdio package it requires a human to deploy it before any agent can reach it, so the
    mode is recorded as local-stdio rather than remote. No endpoint URL is asserted
    because no public one exists.
servers:
- name: lscg (context-graph)
  vendor_unit: Publicis Sapient — Sapient AI Products
  package: '@psnext/lscg'
  version: 0.1.6
  published: '2026-08-05'
  transport: stdio
  start_command: lscg mcp
  auth: none
  source: https://registry.npmjs.org/@psnext/lscg
  method: searched
  tools_source: >-
    The package README's "MCP tools" section. Tool names are verbatim; input schemas were
    not captured because that requires running the server and issuing tools/list against
    it, which needs a local install.
  tools:
  - name: context_graph_scan
    description: Scan the current repository into the repo-local and/or home-level graph databases.
  - name: context_graph_status
    description: Report stored graph state and freshness for a given scope.
  - name: context_graph_nodes
    description: List or search graph nodes, filterable by kind and name.
  - name: context_graph_edges
    description: List graph edges, filterable by kind (calls, imports, exports, attributed_to).
  - name: context_graph_neighbors
    description: Traverse neighbours of a node to a bounded depth.
  - name: context_graph_query
    description: Run a read-only SQL query against the graph database.
  tool_count: 6
- name: gravity-bot-mcp (knowhow-mcp)
  vendor_unit: Publicis Sapient — KnowHOW
  repository: https://github.com/PublicisSapient/knowhow-mcp
  license: Apache-2.0
  transport: http-sse
  protocol_version: '2024-11-05'
  server_version: 1.0.0
  context_path: /mcp/api
  default_port: 8085
  paths:
    sse: /mcp/sse
    messages: /mcp/messages?sessionId={sessionId}
  auth: none
  source: >-
    src/main/java/com/publicissapient/knowhow/knowhow/mcp/McpService.java and
    McpController.java in PublicisSapient/knowhow-mcp
  method: searched
  tools:
  - name: ask_gravity_bot
    description: Ask a question about internal documentation (KPIs, DSR, etc.) using RAG.
    input_schema:
      type: object
      properties:
        question:
          type: string
          description: The question to ask the bot
      required:
      - question
  tool_count: 1
  note: >-
    Retrieval-augmented assistant over the KnowHOW Confluence space
    (knowhow.suite.publicissapient.com/wiki, space key PS). Requires an operator-supplied
    LLM endpoint, a PostgreSQL/pgvector store and Confluence credentials; the shipped
    defaults point at internal Publicis Sapient infrastructure that does not resolve
    publicly.
counts:
  servers: 2
  tools: 7
  remote_endpoints: 0