C3 AI Documentation Home

Configure and Use C3 Central MCP Server

After connecting to the C3 Central MCP Server, you can configure instruction files, enable tools, use prompts, discover skills, and install extensions to create a personalized development workflow.

This setup allows your coding assistant to generate C3 AI code that aligns with your project standards and Type system conventions.

For setup instructions, see Connect to the C3 Central MCP Server.

Set up instruction files

Instruction files provide persistent context about your C3 AI project, such as naming conventions, Type patterns, and file structure. Your coding assistant uses these files to produce accurate and consistent code.

Install instruction files

  1. Open a conversation with your coding assistant.

  2. Enter the command: Update my instructions using C3 MCP.

  3. The assistant calls the get_latest_instructions tool to fetch the latest C3 instructions and installs the c3-mcp-cli command-line tool.

    Instruction setup check

  4. The assistant downloads and runs the setup script automatically.

    Instruction installation in progress

  5. Wait for the script to complete. The instruction files are stored in different locations depending on your editor:

    • VS Code: .github/instructions, .github/skills folders
    • Cursor: .cursor/rules, .cursor/skills folders

    Instruction installation complete

If the command does not run automatically, copy it from the chat and paste it into your terminal (Bash for Windows, standard terminal for Linux or macOS).

Tools

The MCP Server provides tools your coding assistant uses automatically when responding to your requests. You can enable or disable them depending on your workflow.

Available tools

The following table lists the core tools provided by the C3 Central MCP Server. Additional tools may be available depending on your server configuration.

ToolDescription
ask_c3_questionAnswer general knowledge questions in the C3 AI domain.
discover_skillsDiscover and fetch all available skills with their names and descriptions.
edit_c3_typeEdit an existing C3 Type based on a description of the changes.
expand_c3_documentFetch and return the expanded content of a specified C3 developer document.
generate_c3_transformGenerate a JavaScript file for a C3 Transform given source and target Type files.
generate_c3_types_from_dataGenerate C3 Types for multiple data files with proper references between related types.
generate_js_implementationGenerate a JavaScript implementation for a C3 Type based on a description.
generate_new_c3_type_from_descriptionGenerate the content for a new C3 Type based on a description.
generate_python_implementationGenerate a Python implementation for a C3 Type based on a description.
get_latest_instructionsFetch the latest C3 AI coding instructions and install the c3-mcp-cli command-line tool for a specific IDE agent and UI framework.
init_c3_mcpInitialize C3 MCP for a new project or when C3 AI-specific instructions are not present.
search_c3_documentationPerform a semantic similarity search against C3 documentation.
send_feedback_to_c3Send feedback about the AI behavior or tool outputs to C3.

Example tool usage

ActionExample user requestTool usedResult
Search documentation"Show me best practices for creating C3 Transforms."search_c3_documentationReturns relevant documentation and links.
Generate a new Type"Create a C3 Type for a temperature and humidity sensor."generate_new_c3_type_from_descriptionGenerates and formats the new Type.
Modify an existing Type"Add a field called location to the Sensor Type."edit_c3_typeUpdates the Type definition and displays a diff.
Create a Transform"Generate a Transform from RawSensorData to Sensor Type."generate_c3_transformCreates the JavaScript file for the Transform.
Implement logic"Write the implementation for the Sensor Type's validate method."generate_js_implementationGenerates the JavaScript function template.
Generate Python code"Write the Python implementation for the QueryEngine's run method."generate_python_implementationGenerates the Python function implementation.

Enable or disable tools

  1. In your coding assistant chat panel, select the tools icon.
  2. Navigate to MCP Server: C3AI-MCP and review the list of available tools.
  3. Enable or disable tools to match your workflow.

Tool selection window

Only enabled tools are accessible during your sessions.

Test tools

You can test tools directly from the C3 Central MCP Server web portal at https://mcp.c3.ai.

  1. Navigate to the Tools page in the web portal.

  2. Select the play icon in the Actions column for the tool you want to test.

  3. A dialog opens showing the tool description and its required arguments.

  4. Enter test values for each argument and select Run Test.

    Tool testing dialog

This helps you verify tool behavior before using it in your coding assistant.

Prompts

Prompts are predefined workflows that guide your coding assistant through common C3 AI development tasks. You can access them directly from your coding assistant's chat interface.

Use prompts in your editor

  1. In your coding assistant's chat, type /.

  2. Browse or search for a prompt.

  3. Select it to insert it into your chat.

  4. Follow the guidance that appears.

Type the name of a prompt after / (for example, /Create calculated fields in C3) to find it faster.

Available prompts

PromptDescription
Use annotations in C3Provides guidance for defining and using C3 annotations.
Design C3 data models and application architectureOutlines data modeling patterns and application design principles.
Create calculated fields in C3Walks through creating calculated fields following C3 best practices.
Create CronJobs in C3Guides through creating CronJobs in the C3 system.
Write documentation for C3 AI TypesGenerates structured .c3typ documentation templates.
Use the C3 file systemExplains how to manage and navigate files in C3 AI projects.
Write C3 unit tests using JasmineCreates Jasmine-based unit tests for C3 Types and logic.
Set up React applications with C3 AI ReactGuides through React integration setup for C3 AI packages.
Write C3 unit tests using pytestGenerates pytest-based unit tests for backend logic.
Create roles and permissionsProvides steps for defining application roles, permissions, and access control.

Each prompt launches a guided workflow. The assistant asks clarifying questions, then generates code or configuration aligned with C3 AI standards.

Common prompts

TaskExample prompt
Create calculated fields/Create calculated fields in C3
Create CronJobs/Create CronJobs in C3
Set up React applications/Set up React applications with C3 AI React
Write unit tests/Write Jasmine unit tests for C3 Types
Create roles and permissions/Create roles and permissions

Each prompt launches a guided workflow. The assistant asks clarifying questions and generates C3-compliant code or configuration.

Skills

Skills are reusable capabilities that provide specialized knowledge and workflows to AI agents. The C3 Central MCP Server includes a library of skills that your coding assistant can discover and use automatically.

About skills

Skills extend your coding assistant with domain-specific expertise. Each skill contains tested instructions for a particular area of C3 development. You can use the discover_skills tool to list all available skills and their descriptions. Built-in skills are available automatically. To install additional or custom skills, use c3-mcp-cli add-skill <skill-name> (see Install a skill).

View available skills

You can browse all available skills in the C3 Central MCP Server web portal.

  1. Go to https://mcp.c3.ai and navigate to the Skills page.

  2. Search or filter skills by name, description, or status.

    Skills management page

The following table shows some example skills:

SkillDescription
c3-type-generationGenerate C3 Types using the C3 AI MCP tools.
c3-workflowBuild, troubleshoot, and decide when to use C3 Workflows (DAGs).
custom-c3-vault-encryptionEnable, disable, and rotate encryption for the C3 AI Vault (secrets store).
custom-c3-marp-slidesCreate Marp slide decks using C3 AI branding style.
custom-c3-jupyterhubC3 AI data scientist workflow: setup credentials, JupyterHub, and resync packages.
custom-c3-cve-scanfixSearch for CVEs and fixes in third-party OSS dependencies.
custom-c3-app-keys-tokensCreate and manage C3 app keys, secret keys, session tokens, and app tokens.
custom-c3-remotion-videoCreate professional sales demo videos for C3 AI applications using Remotion.

Skills are versioned and managed centrally. New skills are added regularly. Check the Skills page for the latest list.

Install a skill

You can install any available skill to your local workspace using the c3-mcp-cli command-line tool.

  1. Find the skill you want on the Skills page at https://mcp.c3.ai.

  2. Run the following command in your terminal:

    Command Line
    c3-mcp-cli add-skill <skill-name>

    For example:

    Command Line
    c3-mcp-cli add-skill custom-c3-cve-scanfix

  3. The skill is downloaded to your personal skills folder (see Skill storage locations below).

Create custom skills

You can create your own skills to package domain-specific knowledge and workflows for your coding assistant. Each skill consists of a folder containing a SKILL.md file with YAML frontmatter and markdown instructions.

Skill file structure

A minimal custom skill requires a single SKILL.md file inside a named folder:

Text
my-custom-skill/
└── SKILL.md

For more complex skills, you can include additional reference files:

Text
my-custom-skill/
├── SKILL.md
├── reference.md
└── examples/

SKILL.md format

The SKILL.md file uses YAML frontmatter followed by markdown content:

YAML
---
name: my-custom-skill
description: >
  Short description of what the skill does and when to use it.
  Include trigger phrases so the assistant knows when to apply this skill.
---

# My Custom Skill

## Overview

Detailed instructions, patterns, and examples for the assistant to follow.

The description field is important because your coding assistant uses it to determine when the skill applies to your request.

Skill storage locations

Where you store your skill folder depends on your editor and whether you want to share the skill with your team.

LocationEditorScope
.github/skills/VS Code (Copilot)Personal
.cursor/skills/CursorPersonal
.claude/skills/Claude CodePersonal
.agents/skills/All editorsShared with team

Whether a folder is tracked in Git depends on your repository's .gitignore configuration, not on the folder name.

Personal skills are stored in editor-specific folders that are gitignored. Use these folders for experimental or role-specific skills that do not need team review.

Shared skills are stored in the .agents/skills/ folder and can be made available to all contributors on the repository.

Extensions

Extensions allow you to install additional MCP servers alongside the C3 MCP Server. Each extension connects your coding assistant to a third-party service, expanding its capabilities beyond C3 AI development.

Available extensions

You can browse and install extensions from the Extensions page in the C3 Central MCP Server web portal.

Extensions page

The following extensions are available:

ExtensionDescriptionCapabilities
AtlassianAccess Jira, Confluence, and Compass through Atlassian's official MCP server.Search and manage Jira issues, sprints, and boards. Read and create Confluence pages and spaces. Query Compass components and dependencies.
GitHubGitHub's official MCP server for repos, issues, PRs, Actions, and code security.Search code, repositories, and commits. Create and review pull requests. Manage issues, discussions, and CI/CD workflows.
FigmaGenerate code from Figma designs using Figma's official remote MCP server.Generate code from selected frames and components. Extract variables, styles, and layout data. Use Code Connect to keep output consistent with your codebase.
LucidSearch, summarize, share, and create Lucid diagrams from your IDE.Create flowcharts, org charts, sequence diagrams, and more from data. Share documents and create shareable links with collaborators.
Chrome DevToolsDebug and automate Chrome from your IDE using Google's official DevTools MCP server.Inspect network requests, console logs, and performance traces. Automate browser actions such as navigate, select, and fill forms. Take screenshots and verify fixes in real time.
PlaywrightBrowser automation and testing powered by Microsoft's official Playwright MCP server.Navigate pages, select, fill forms, and take screenshots. Run end-to-end tests from natural language. Uses accessibility tree with no vision model needed.

Install an extension

  1. Go to https://mcp.c3.ai and navigate to the Extensions page.
  2. Locate the extension you want to install.
  3. If the extension requires credentials (such as a GitHub Personal Access Token), enter them in the provided field.
  4. Select VS Code, Cursor, or Claude to install the extension in your editor.

All credentials are stored locally in your browser and are never sent to the C3 server.

Connect to a C3 AI application

You can connect the MCP Server to a Single Node Environment (SNE) to work with live Types and configurations from your C3 AI application.

  1. Open the C3 AI VS Code Extension.

  2. Connect to your SNE from the extension settings.

  3. The MCP Server detects the connection automatically.

  4. Your coding assistant can now:

    • Access live Type definitions.
    • Generate Types and Transforms consistent with your data model.
    • Suggest code that matches your application's configuration.

If the assistant suggests incorrect fields or relationships, confirm that the SNE connection points to the correct environment.

Troubleshoot configuration issues

Instruction files not detected

Problem: Copilot does not reference instruction files after installation.

Solution:

  1. Verify the .github folder exists at your project root level.

  2. Check the folder structure:

    Text
    .github/
├── instructions/
│   └── copilot-instructions.md
└── workflows/

  3. If instruction files are in a nested .github folder, move them to the root-level .github directory.

  4. Restart VS Code and check for the "Used 1 reference" indicator in Copilot Chat.

Multiple .github folders can cause confusion. Always use the top-level directory for MCP instruction files.

MCP server not responding

Problem: MCP tools appear in the prompt list but do not execute.

Solution:

  1. Open Command Palette (Cmd+Shift+P on macOS, Ctrl+Shift+P on Windows/Linux).
  2. Select MCP: Open User Configuration.
  3. Check the status indicator for each server entry.
  4. If status shows anything other than "Running", click Start to activate the server.
  5. Verify all server entries show "Running" status.

The MCP server does not start automatically. You must manually activate it after installation or when switching projects.

Code changes not applying automatically

Problem: Copilot suggests code but does not modify files directly.

Solution:

  1. Check the mode selector at the bottom of Copilot Chat.
  2. Click the dropdown (may show "Claude Sonnet 4" or another model).
  3. Select Agent from the options.
  4. Verify Agent mode is active by confirming:
    • "Agent" appears in the mode selector
    • Context indicators show active extensions
    • The prompt shows "Add context (#), extensions (@), commands (/)"

Agent mode is required for automatic file modifications. Without it, you only receive code suggestions.

Tools not available after configuration

Problem: Enabled tools do not appear or respond in chat.

Solution:

  1. Verify the MCP server is running (see "MCP server not responding" above).
  2. Check that Agent mode is enabled (see "Code changes not applying automatically" above).
  3. Type "/" in Copilot Chat to view available prompts.
  4. If MCP prompts appear but tools don't work, restart VS Code and repeat verification.

After resolving issues, test with: "What tools do you have access to?" to confirm connectivity.

See also

Was this page helpful?