tradingrearchagents
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| TRADINGAGENTS_LEGACY_PATH | No | Path to a compatible TradingAgents checkout. Takes precedence over the pinned upstream extra. Set when using the legacy upstream execution path. | |
| TRADINGAGENTS_CODEX_AUTH_PATH | No | Optional path to Codex OAuth auth.json (default ~/.codex/auth.json). Used only with the openai_codex provider on a checkout containing TradingAgents PR #1195. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| discover_capabilityA | Discover executors, tools, safety boundaries, and default fixture. |
| get_feature_matrixA | Return implemented features, safety exclusions, and runtime readiness. |
| prepare_fixtureA | Validate and expand the deterministic ORCL fixture without running it. |
| run_fixtureB | Run all stages deterministically and store the in-memory result and events. |
| prepare_host_runB | Return the canonical plan for the active host harness; accepts no model credentials. |
| import_host_runB | Validate completed host stage outputs and atomically publish the final dossier. |
| create_host_runC | Create a durable credential-free host run with decision-memory recall. |
| start_host_runA | Start a prepared durable host run and return its first stage. |
| append_run_receiptsC | Append safe live stage/tool receipts without prompts, raw arguments, or credentials. |
| commit_host_stageA | Commit and checkpoint one completed host stage, then return the next stage. |
| pause_host_runC | Pause a host run at its next portable stage boundary. |
| resume_host_runB | Resume from the first incomplete portable stage; replay interrupted work. |
| get_run_controlB | Return durable lifecycle status, revision, checkpoint, cancellation, and next stage. |
| poll_run_eventsC | Read live lifecycle events after a monotonic cursor. |
| request_run_cancellationB | Request cooperative cancellation; the host owns interruption. |
| acknowledge_run_cancellationC | Acknowledge host interruption and make cancellation terminal. |
| finalize_host_runA | Validate all committed stages and atomically publish the completed dossier. |
| export_completed_runA | Export a bundle with atomic first publication and crash-recoverable validated overwrite. |
| query_decision_memoryA | Recall up to five same-symbol and three cross-symbol prior decisions. |
| record_decision_outcomeB | Append a host-observed outcome and reflection to a prior research decision. |
| get_conformance_reportB | Validate portable observable invariants; optionally verify pinned checkout identity. |
| get_runA | Return a compact dashboard-oriented run record. |
| get_run_eventsB | Return the ordered event stream for a run. |
| get_run_resultB | Return the full typed result for a run. |
| get_run_viewA | Return the complete UI-ready run projection for inline harness rendering. |
| launch_local_dashboardC | Serve the local dashboard assets and JSON APIs on loopback. |
| get_dashboard_reportB | Return the presentation-safe dashboard summary for a run. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 27 tools
Several tools have overlapping boundaries: get_run, get_run_events, get_run_result, get_run_view, get_run_control all retrieve run data but differ subtly in scope and purpose. Similarly, run_fixture, create_host_run, and start_host_run all initiate executions, and import_host_run vs finalize_host_run both validate and publish dossiers. The differences are hard to discern from names alone.
Most tools follow a consistent verb_noun pattern (get_, create_, start_, run_, pause_, resume_, finalize_), with a few camelCase outliers like launch_local_dashboard, prepare_host_run, and append_run_receipts that break the underscore convention. The dominant pattern is readable and predictable, but the mixed style is a minor inconsistency.
27 tools is on the heavy side, especially for what appears to be a single-purpose trading research harness. Many tools (get_run, get_run_events, get_run_result, get_run_view, get_run_control, poll_run_events) overlap substantially in the retrieval space, suggesting the surface could be consolidated significantly.
The host-run lifecycle is comprehensively covered: create, start, commit, pause, resume, cancel, finalize, export, and query. Decision memory has both recall and record-outcome operations. Minor gaps exist around fixture editing or host-run deletion, but the core research workflow appears fully covered with no obvious dead ends.