Know when a strategy should not ship.
Test a strategy before an agent deploys or scales it. AlphaLitmus challenges costs, delays, regimes, parameters, and source evidence—then returns a verified, reproducible release decision.
Import ChallengeRequest JSON
Supply the complete request, not a Report or bare candle array. Reference mode requires research with daily candles, source and data_kind. The server validates the contract. Never include credentials. Maximum import: 2 MB.
Nexus evidence challenge (explicit mode: nexus)
Read-only server-side evidence request. Credential availability is unknown to this page; missing access or evidence is reported, not treated as validation. No API keys belong in this form.
Run the safe release-gate demo to generate a bounded, reproducible decision.
Why this decision
Awaiting current release-gate result.
Observed failures
- Awaiting current result; no failures shown yet.
Unavailable evidence
Report identity
- Report ID
- Unavailable
- Canonical hash
- Unavailable
Public synthetic demo: decision is INSUFFICIENT_EVIDENCE. Synthetic evidence cannot approve or reject a real production strategy. Observed failures remain visible; more authentic evidence is required.
This result is not financial advice and is not deployment approval.
Inspect complete release-gate JSON
Unavailable / no current result.
Fetch four read-only Nexus evidence surfaces for the fixed BTC/USDT candidate and run the verified release gate.
Live evidence details Identity, freshness, failures, and export
A live pulse is evidence freshness, not trading authorization, source authentication, or proof of future profitability.
Fixed identity and freshness
- Strategy
- Unavailable
- Strategy ID
- Unavailable
- Symbol
- Unavailable
- Upstream fetched
- Unavailable
- Surfaces
- Unavailable
- Freshness
- Unavailable
- Cache
- Unavailable
- Report ID
- Unavailable
Observed failures
- Awaiting live check.
Inspect complete live result JSON
Unavailable / no live result yet.
Replay recorded historical evidence through the production reconciliation engine. No live call, no key required.
Recorded evidence details Identity, integrity, decision, and limitations
Recorded historical Nexus evidence — not a live market call and not evidence of future profitability.
Evidence identity
- Strategy
- Unavailable
- Strategy ID
- Unavailable
- Run ID
- Unavailable
- Capture time
- Unavailable
- Surfaces
- Unavailable
- Integrity
- Unavailable
- Report ID
- Unavailable
Release decision
- Awaiting recorded replay; no result shown yet.
TRADE_SYMBOLS
Unavailable
Limitations
- Awaiting recorded replay.
Inspect complete recorded replay JSON
Unavailable / no recorded replay yet.
When should an agent call this?
- before enabling a strategy
- after changing strategy configuration
- before increasing risk or scale
- after a market-regime change
- during scheduled strategy revalidation
Agent call contract / copyable examples
One strict release-gate contract across REST and MCP, with verified source reports and bounded decisions.
View REST, MCP, and response examples
REST call
POST /v1/release-gate
Content-Type: application/json
{"mode":"nexus","symbol":"BTC/USDT"}
MCP tool arguments
{"name":"evaluate_strategy_release","arguments":{"request":{"mode":"nexus","symbol":"BTC/USDT"}}}
Expected decision fields
{
"schema_version": "alphalitmus-release-gate-1",
"decision": "BLOCK_DEPLOYMENT | INSUFFICIENT_EVIDENCE | SURVIVED_BOUNDED_TESTS",
"recommended_action": "DO_NOT_DEPLOY | COLLECT_MORE_EVIDENCE | CONTINUE_PAPER_VALIDATION",
"source_verdict": "FRAGILE | INCONSISTENT | UNPROVEN | SURVIVED_TESTS",
"report_id": "<64-hex>",
"canonical_report_hash": "<64-hex>",
"no_execution": true,
"profitability_claimed": false
}
Primary MCP tool: evaluate_strategy_release. Existing tools remain available. SURVIVED_BOUNDED_TESTS means only survival of the stated bounded tests, not deployment approval.
Run the judge demo to generate a bounded, evidence-labeled result.
Zero-centered; guarded variant only.
Additional modeled bps, not a forecast.
Local neighborhood, not optimization advice.
Deep evidence report11 reproducible sections · metrics, stress tests, certificate, and provenance
01 / Strategy & evidence
02 / Verdict
03 / Recomputed metrics
04 / Claims vs recomputed
05 / Stress matrix
06 / Fragility frontier
07 / Minimal failure boundary
08 / Certificate
09 / Provenance & hashes
10 / Limitations & unavailable evidence
11 / Export & usage: REST / MCP
Export is enabled only for the latest successful Report. Starting any new request clears the previous report and disables export. The JSON includes the submitted request and may contain your dataset; review before sharing.
Inspect complete current Report JSON
Unavailable / no current report.
REST / same-origin service
GET /v1/release-gate/demo/mixed
GET /v1/release-gate/demo/shock
POST /v1/release-gate
Content-Type: application/json
{"mode":"nexus","symbol":"BTC/USDT"}
# Primary release-gate call (same ChallengeRequest as /v1/challenge):
curl -X POST http://127.0.0.1:8000/v1/release-gate -H "Content-Type: application/json" --data-binary @challenge.json
# Legacy detail routes remain available:
# GET /v1/demo/mixed, GET /v1/demo/shock, POST /v1/challenge
MCP / local stdio
python -m app.mcp_server
# Primary tool:
# {"name":"evaluate_strategy_release","arguments":{"request":{"mode":"nexus","symbol":"BTC/USDT"}}}
Connect your MCP client to that local command, run tools/list, and use the advertised tool schema. Primary tool is evaluate_strategy_release; the existing five tools remain available. MCP connectivity has not been checked by this page.
All page assets are local. Demo and import requests go to this origin; explicit Nexus mode may cause the server to contact Nexus. Opening this HTML as a file is not sufficient: serve it through the application.