Tools and Workflows

Professional

Algo Trade Analytics MCP exposes a research toolpack, not a single chat command. The client owns the reasoning loop. Algo Trade Analytics owns the trading-specific state, tools, gates, and evidence records.

Tool Families

Family	Main tools	Use for
Context and status	ata_health, ata_context, ata_mcp_usage_status	Confirm server health, bound actor, entitlement, market-data readiness, and usage status.
Guidance resources	ata_list_resources, ata_read_resource, ata_list_prompts, ata_get_prompt	Load the Algo Trade Analytics Quant Research Driver skill, prompt templates, and tool manifest.
Case Studies	ata_list_case_studies, ata_load_case_study, load_case_study, create_research_run	Find the user’s research ledger and bind tool calls to the correct investigation.
Notebook and plan	ata_get_research_notebook, ata_get_research_plan, ata_plan_case_study_hardening	Inspect evidence maturity, missing gates, and recommended next methodology tasks.
Research contract	ata_start_research_run, ata_get_research_contract, ata_research_workflow_status	Keep the external agent aligned with the current allowed tools, blockers, and handoff state.
Light analysis	inspect_market_context, analyze_losing_trades, analyze_trade_features, analyze_market_coverage, analyze_signal_audit, propose_strategy_candidate, analyze_parameter_robustness	Explore market context, failures, features, signals, and candidate ideas before expensive validation.
Heavy evidence	profile_baseline, evaluate_candidate, verify_candidate, stress_test_candidate, walk_forward_candidate, run_parameter_sweep, analyze_bootstrap_confidence, analyze_overfit_probability, analyze_cost_sensitivity, analyze_risk_management, analyze_cross_scope	Run paid evidence checks, validation, sweeps, stress tests, robustness analysis, and cross-scope checks.
Background jobs and tasks	ata_start_research_job, ata_get_research_job_status, ata_cancel_research_job, ata_get_research_task, ata_get_research_task_result, ata_cancel_research_task, ata_list_research_tasks	Run heavy work durably instead of keeping one client call open.
Platform workflows	ata_start_research_workflow, ata_get_research_workflow_status, ata_cancel_research_workflow, ata_resume_research_workflow	Start and manage a platform-managed research workflow from a higher-level objective.
Write tools	create_research_hypothesis, define_experiment_protocol, close_research_experiment, stage_candidate_for_editor, commit_research_lesson, ata_record_research_task_progress	Mutate the research ledger, stage candidates, or save lessons when write scopes are granted.

Recommended Call Order

1. Call ata_context first. Confirm the Algo Trade Analytics account, MCP status, market-data readiness, and client guidance.

2. Load guidance. If the client did not load MCP resources automatically, call ata_read_resource for skill://ata-quant-research-driver/SKILL.md or call ata_get_prompt with ata_quant_research_driver.

3. Choose the Case Study. Use ata_list_case_studies and ata_load_case_study.

4. Check the contract. Use ata_get_research_contract or ata_research_workflow_status before paid or write tools.

5. Run cheap context tools before heavy validation. Inspect failures, market context, and candidate ideas before spending credits on sweeps or verification.

6. Use jobs for heavy tools. Start a background job, poll status, and read the final result when complete.

7. Write only after evidence exists. Record progress, define protocols, stage candidates, or commit lessons only when the tool result supports it.

Tip

After every tool call, read the returned meta.clientGuidance and meta.researchState. Those fields tell the client what changed, what is blocked, and what should happen next.

Durable Background Jobs

Use background jobs when a tool may run multiple trials, consume a budget, or take long enough that a single MCP request could time out.

Typical job-backed work:

• parameter sweeps,
• walk-forward checks,
• bootstrap confidence checks,
• overfit probability checks,
• cross-scope validation,
• risk-management analysis,
• platform-managed research workflows.

Queued jobs can be cancelled before work starts. Running jobs stop at the next safe trial checkpoint and settle based on completed work.

Candidate Handoff

MCP can stage a candidate for review, but it does not silently replace your active strategy.

1. The external agent proposes or verifies a candidate.
2. A write-scoped tool stages the candidate into Algo Trade Analytics.
3. Algo Lab shows the staged diff or candidate state.
4. You review, accept, reject, or load the source in the app.

This keeps local-agent autonomy separate from final source changes.

Related

Case Study Tab Where MCP evidence and candidate state are stored

Security, Scopes, and Credits Which scopes are required and how Algo Trade Analytics credits apply

Algo Agent Compare MCP-driven research with Algo Trade Analytics's native agent loop