Skip to content

Repository files navigation

FXMacroData for LibreChat

Use your FXMacroData subscription in LibreChat research conversations to compare covered currencies, explore full available indicator histories and analyse release calendars. Native tools and Skills return source-linked tables and complete structured responses for your research.

Subscribe to FXMacroData for access to covered non-USD datasets and full available history.

Evaluate the plugin before subscribing with public USD catalogue, recent history (currently a rolling 90-day window) and release calendars, which require no FXMacroData account or key.

The package exposes all 23 REST operations and 49 MCP tools in its public discovery snapshot.

FXMacroData · API reference

Install

LibreChat must support the Agent Plugins 1.0.0 format. Install this package in the Python environment available to the LibreChat API process (Python 3.10 or later):

python -m pip install .
fxmacrodata-librechat --install-plugin /path/to/LibreChat/plugin

Use your configured DEPLOYMENT_PLUGINS_DIR when it differs from plugin. The installer creates a new fxmacrodata child directory and refuses to overwrite an existing plugin. Restart LibreChat after installation. Its API process must have the installed fxmacrodata-librechat executable on PATH; container deployments must install the wheel inside that container, not only on the host.

Enable the FXMacroData tools and the fxmacrodata-macro-brief or fxmacrodata-series-research deployment Skill in the normal Agent/Skills interface. Ask for a USD macro briefing or a source-linked release calendar. The tools are also available through LibreChat's MCP tool picker. Normal LibreChat model configuration is independent of data access.

The default plugin uses public access without reading credentials. Connect your subscription below to use protected datasets.

Connect your subscription

Merge the server entry from authenticated.librechat.yaml into your existing operator configuration, then restart LibreChat. Enter your key through its sensitive MCP variable setting. This overrides only the default anonymous server connection; the installed tools and Skills remain available. Do not put a key in the YAML, plugin files, prompts, command arguments or source control. Each authenticated connection receives its user's credential through the host's runtime environment.

Use the optional YAML configuration for per-user credentials; LibreChat's plugin mcp.json supports path expansion only.

Research output

Each tool returns a readable table preview and structuredContent containing the original response, its tabular projection, source URL and provider link. All rows and response metadata remain in structuredContent. No units, timestamps, forecast classes or source labels are inferred. Event streams are finite captures bounded by duration and count.

Website links contain static attribution tags. API requests, MCP requests and canonical provenance URLs are untagged. The plugin sends no click beacons.

Development

python -m pip install -e . pytest pytest-asyncio build
python -m pytest tests -q
python -m build

The adapter vendors the MIT-licensed FXMacroData public client. It connects only to documented FXMacroData services. See CAPABILITIES.md for the operation matrix and PUBLIC_FILES.json for the source allowlist.

Releases

Packages

Used by

Contributors

Languages