io.github.cyanheads

evals-mcp-server

Author verifiable eval records through a draft→review→revise→submit loop with enforced graders.

communityapplication

Package Details

@cyanheads/evals-mcp-server

0.1.2•https://www.npmjs.com/package/@cyanheads/evals-mcp-server
Transportstdio
Runtimenode

Package Arguments

Positional
Position 0
run
Position 1
start:stdio

Environment Variables

EVALS_DATA_DIR(str)
Default:./evals-data

Root folder for record JSON. The store manages drafts/, submitted/, and exports/ subdirs under it.

EVALS_REQUIRE_CONFIRMATION(str)
Default:false

When 'true', evals_submit_draft requests a human confirmation where the client supports elicitation.

EVALS_DEFAULT_LICENSE(str)

Default metadata.license applied when a draft omits one.

EVALS_CAPTURE_DIR(str)

Directory of framework-written tool-call captures; when set, captures EvalsIDs resolve to full dumps.

MCP_LOG_LEVEL(str)
Default:info

Sets the minimum log level for output (e.g., 'debug', 'info', 'warn').

@cyanheads/evals-mcp-server

0.1.2•https://www.npmjs.com/package/@cyanheads/evals-mcp-server
TransportStreamable HTTP
Runtimenode

Package Arguments

Positional
Position 0
run
Position 1
start:http

Environment Variables

MCP_HTTP_HOST(str)
Default:127.0.0.1

The hostname for the HTTP server.

MCP_HTTP_PORT(str)
Default:3010

The port to run the HTTP server on.

MCP_HTTP_ENDPOINT_PATH(str)
Default:/mcp

The endpoint path for the MCP server.

MCP_AUTH_MODE(str)
Default:none

Authentication mode to use: 'none', 'jwt', or 'oauth'.

EVALS_DATA_DIR(str)
Default:./evals-data

Root folder for record JSON. The store manages drafts/, submitted/, and exports/ subdirs under it.

EVALS_REQUIRE_CONFIRMATION(str)
Default:false

When 'true', evals_submit_draft requests a human confirmation where the client supports elicitation.

EVALS_DEFAULT_LICENSE(str)

Default metadata.license applied when a draft omits one.

EVALS_CAPTURE_DIR(str)

Directory of framework-written tool-call captures; when set, captures EvalsIDs resolve to full dumps.

MCP_LOG_LEVEL(str)
Default:info

Sets the minimum log level for output (e.g., 'debug', 'info', 'warn').