MCP server

The CSPR.cloud MCP server lets AI agents connect to CSPR.cloud through the Model Context Protocol. Use it when your agent supports remote MCP servers and you want it to call CSPR.cloud tools directly.

Community project

The hosted MCP server is based on the community Casper MCP server. The source repository is available at github.com/msanlisavas/casper-mcp.

Endpoints

Network	MCP endpoint
Mainnet	https://mcp.cspr.cloud/mcp
Testnet	https://mcp.testnet.cspr.cloud/mcp

API key

The hosted MCP server requires a CSPR.cloud API key. Store it in an environment variable and do not commit it to source control.

export CSPR_CLOUD_API_KEY="your-api-key"

Codex setup

Codex stores MCP server configuration in ~/.codex/config.toml. You can add the server with the CLI and then update the configuration to pass the API key as an HTTP header.

1. Add the MCP server

For Mainnet:

codex mcp add cspr_cloud --url https://mcp.cspr.cloud/mcp

For Testnet:

codex mcp add cspr_cloud_testnet --url https://mcp.testnet.cspr.cloud/mcp

2. Add the API-key header

Open ~/.codex/config.toml and add env_http_headers to the CSPR.cloud MCP server entry.

Mainnet:

[mcp_servers.cspr_cloud]
url = "https://mcp.cspr.cloud/mcp"
env_http_headers = { "X-CSPR-Cloud-Api-Key" = "CSPR_CLOUD_API_KEY" }

Testnet:

[mcp_servers.cspr_cloud_testnet]
url = "https://mcp.testnet.cspr.cloud/mcp"
env_http_headers = { "X-CSPR-Cloud-Api-Key" = "CSPR_CLOUD_API_KEY" }

env_http_headers maps the HTTP header name to the environment variable that contains the value. In the example above, Codex reads CSPR_CLOUD_API_KEY and sends it as X-CSPR-Cloud-Api-Key.

3. Verify the connection

Start a new Codex session and run:

/mcp

The CSPR.cloud MCP server should appear in the active MCP servers list. If it does not, check that CSPR_CLOUD_API_KEY is exported in the shell where Codex is started and that the server entry is enabled in ~/.codex/config.toml.

Claude Code setup

Claude Code supports remote MCP servers over HTTP. Use claude mcp add-json so the API key can stay in the CSPR_CLOUD_API_KEY environment variable instead of being written directly into Claude configuration.

Claude Code supports multiple MCP scopes:

• local - available only in the current project. This is the default.

• project - shared through a .mcp.json file in the project root.

• user - available across all projects for the current user.

The examples below use --scope project, which writes a shareable .mcp.json file with an environment-variable placeholder instead of the API key. Claude Code asks you to approve project-scoped MCP servers before using them.

1. Add the MCP server

For Mainnet:

claude mcp add-json cspr_cloud --scope project \
  '{"type":"http","url":"https://mcp.cspr.cloud/mcp","headers":{"X-CSPR-Cloud-Api-Key":"${CSPR_CLOUD_API_KEY}"}}'

For Testnet:

claude mcp add-json cspr_cloud_testnet --scope project \
  '{"type":"http","url":"https://mcp.testnet.cspr.cloud/mcp","headers":{"X-CSPR-Cloud-Api-Key":"${CSPR_CLOUD_API_KEY}"}}'

Claude Code expands ${CSPR_CLOUD_API_KEY} from the environment when it loads the MCP server. Start Claude Code from a shell where the variable is exported.

2. Verify the connection

From a terminal:

claude mcp list

Inside Claude Code:

/mcp

The CSPR.cloud MCP server should appear in the MCP servers list. If Claude reports a configuration parsing error, check that CSPR_CLOUD_API_KEY is set in the shell where Claude Code starts.