MCP Integration (Claude Code & Codex)
The Model Context Protocol (MCP) is an open standard that lets AI coding assistants call external tools and services during a conversation. By connecting CurieTech AI as an MCP server, your coding agent gains direct access to integration intelligence without leaving the terminal or IDE you already work in.
CurieTech AI MCP is available for Claude Code and Codex, bringing domain-aware assistance into the agentic development environments teams are adopting every day.
Why MCP Connectivity Matters
Developers spend time switching between their coding environment and external tools to handle integration-specific work. With CurieTech AI as an MCP server, your coding agent can:
- Generate DataWeave transformations on demand, directly within your coding workflow
- Refactor and update existing transformations without copying code back and forth
- Validate transformation logic and get expert feedback inline
- Access integration intelligence tuned to your codebase and domain context
Once connected, invoking CurieTech AI is as natural as any other prompt just tell your coding agent to use CurieTech AI for the task.
Prerequisites
Before setting up the integration, you'll need:
- Claude Code installed and running, or access to Codex
- A CurieTech AI account
- A CurieTech AI API key generated from your account settings
Step 1: Generate an API Key
Regardless of which coding agent you use, start by generating an API key:
- Log in to your CurieTech AI account
- Navigate to Settings → API Key Management
- Click Generate API Key
- Copy the key and store it securely you won't be able to see it again
Connect to Claude Code
Step 2: Register the MCP Server
Run the command below in your terminal. You do not need to provide your API key in the command; Claude Code will open a CurieTech AI authorization page when the MCP server is first used.
# Register CurieTech AI as an MCP server in Claude Code
claude mcp add --transport http curie-dataweave https://platform.curietech.ai/mcp/ --scope user
This registers CurieTech AI as an MCP server named curie-dataweave under your user scope. The --scope user flag makes it available across all your Claude Code projects. Keep the trailing slash in https://platform.curietech.ai/mcp/ for best compatibility with MCP clients.
Step 3: Authorize CurieTech AI
After registration, start Claude Code and ask it to use the CurieTech AI MCP server. Claude Code will detect that the server requires authorization and show a CurieTech AI OAuth authorization URL.
- Open the authorization URL shown by Claude Code
- Paste your CurieTech AI API key into the authorization form
- Click Authorize
- Return to Claude Code after the browser redirects back to Claude's local callback URL
Claude Code stores the OAuth token it receives from CurieTech AI, so you do not need to paste your API key into every prompt. You may need to authorize again if the token expires, is revoked, or Claude Code's local MCP credentials are cleared.
Step 4: Verify the Connection
claude mcp list
You should see curie-dataweave listed as an available MCP server.
Step 5: Invoke CurieTech AI
Once connected, reference the MCP server in any Claude Code prompt:
Generate a new DataWeave transformation:
Use CurieTech AI MCP server to generate a DataWeave transformation that maps the input JSON payload to the target schema.
Refactor an existing transformation:
Use CurieTech AI MCP server to refactor this DataWeave script to handle null values and improve readability.
Validate a transformation:
Use CurieTech AI MCP server to validate this DataWeave transformation and identify any issues.
Claude Code will route the request to CurieTech AI and return the result inline no context switching required.
Connect to Codex
Step 2: Register the MCP Server
Run the command below in your terminal. You do not need to provide your API key in the command; Codex will open a CurieTech AI authorization page when the MCP server is first used.
# Register CurieTech AI as an MCP server in Codex
codex mcp add curie-dataweave --url https://platform.curietech.ai/mcp/
This registers CurieTech AI as an MCP server named curie-dataweave in Codex. Keep the trailing slash in https://platform.curietech.ai/mcp/ for best compatibility with MCP clients.
Step 3: Authorize CurieTech AI
After registration, start Codex and ask it to use the CurieTech AI MCP server. Codex will detect that the server requires authorization and show a CurieTech AI OAuth authorization URL.
- Open the authorization URL shown by Codex
- Paste your CurieTech AI API key into the authorization form
- Click Authorize
- Return to Codex after the browser redirects back to Codex's local callback URL
Codex stores the OAuth token it receives from CurieTech AI, so you do not need to paste your API key into every prompt. You may need to authorize again if the token expires, is revoked, or Codex's local MCP credentials are cleared.
Step 4: Verify the Connection
Open Codex and check that curie-dataweave appears in your list of configured MCP servers.
Step 5: Invoke CurieTech AI
Once connected, reference the MCP server in any Codex prompt:
Generate a new DataWeave transformation:
Use CurieTech AI MCP server to generate a DataWeave transformation that maps the input JSON payload to the target schema.
Refactor an existing transformation:
Use CurieTech AI MCP server to refactor this DataWeave script to handle null values and improve readability.
Validate a transformation:
Use CurieTech AI MCP server to validate this DataWeave transformation and identify any issues.
Codex will route the request to CurieTech AI and return the result inline no context switching required.
Benefits
- No context switching: Integration intelligence lives inside your existing coding workflow
- Domain-aware assistance: CurieTech AI understands your integration codebase, not just generic code patterns
- Faster delivery: Generate, refactor, and validate transformations in seconds from your terminal
- Works where you work: Compatible with Claude Code and Codex the tools your team already uses
Troubleshooting
command not found when running claude mcp add
Ensure Claude Code is installed and the claude binary is on your PATH. See the Claude Code documentation for installation steps.
command not found when running codex mcp add
Ensure Codex is installed and available on your PATH. Refer to the Codex documentation for installation steps.
Claude Code shows an authorization URL This is expected. Open the URL, paste your CurieTech AI API key into the authorization form, and click Authorize. Claude Code will exchange the authorization response for an MCP access token.
Coding agent still cannot connect after authorization Your API key may be incorrect or expired. Re-generate a key from Settings → API Key Management, then retry the authorization flow.
Codex shows an authorization URL This is expected. Open the URL, paste your CurieTech AI API key into the authorization form, and click Authorize. Codex will exchange the authorization response for an MCP access token.
MCP server not appearing after registration (Claude Code)
Restart Claude Code after registering the server and run claude mcp list again.
MCP server not appearing after registration (Codex) Restart Codex after registering the server and check your configured MCP servers again.
For further help, visit Troubleshooting & Support.
