Developer Platform

Build on
quantitative
consensus.

45+ quantitative models vote nightly on US equities. Access the consensus via REST API, connect an LLM agent via MCP, or wire your own agent directly via A2A protocol.

signa — api query
# Get latest Tier 3 signals
$ curl "https://api.getsigna.ai/v1/signals?tier=3" \
  -H "Authorization: Bearer sk-signa-..."
 
↳ 9:05 PM ET nightly · 45+ models voted
 
"ticker": "NVDA"
"direction": BULLISH  TIER 3
"composite_score": 91.4
"confidence": 0.87
"model_count": 14
"key_drivers": [
  "Stage 2 advance · 14.9% above 30w MA",
  "Data center +190% YoY · accelerating",
  "EPS beat streak · 3 consecutive qtrs"
]
_ |
45+
Quant agents
8×
MCP tools
3
Integration surfaces
9pm
Nightly pipeline ET
01
⚡
REST API
Standard HTTP/JSON endpoint. Pull signals, run screener queries, fetch fundamentals and news sentiment programmatically.
Live
For developers, quant tools, dashboards
02
🤖
MCP Server
Connect Claude, GPT, or Cursor to Signa via Model Context Protocol. Ask signal questions in plain language.
Beta
For AI assistants, research agents
03
🔗
A2A Protocol
Agent-to-agent integration with immutable attribution, confidence intervals, and chain-of-custody metadata baked in.
Beta
For AI pipelines, multi-agent systems
01 — REST API

Standard HTTP · JSON · Bearer auth

Every endpoint returns a structured JSON response with a consistent root schema. Your first call should take under 5 minutes.

Fetch the latest Tier 3 signals

# Tier 3 = highest conviction, 4+ model categories agree
curl -X GET "https://api.getsigna.ai/v1/signals?tier=3&limit=10" \
  -H "Authorization: Bearer sk-signa-prod-YOUR_KEY" \
  -H "Content-Type: application/json"

Understanding the response

Response
{
  "signals": [{
    "ticker":           "NVDA",        ← A
    "direction":        "BULLISH",
    "alert_tier":       3,
    "composite_score":  91.4,         ← B
    "confidence":       0.87,          ← C
    "model_count":      14,
    "regime":           "RISK_ON",
    "key_drivers":      ["..."],       ← D
    "risks":            ["..."],
    "expires_at":       "2026-04-07T..."
  }]
}
A
ticker
The equity symbol. All signals are single-ticker — no basket or index signals.
B
composite_score
0–100 weighted score. Factors: model confidence (50%), model count (25%), category diversity (25%). Regime-gated.
C
confidence
0–1 probability-like estimate. Regime-adjusted weighted average of all contributing model confidences.
D
key_drivers
Top 3 reasons the signal fired. Human-readable. Same strings shown in the Action Card.
02 — MCP SERVER

Ask signal questions
in plain language.

Connect Claude, GPT, or Cursor to Signa via Model Context Protocol. Your AI assistant gets 8 tools — signals, fundamentals, news, options flow, screener, regime, leaderboard, and company profile.

01
Copy the MCP URL
https://api.getsigna.ai/mcp/sse

This is your SSE endpoint. Works with any MCP-compatible client.
02
Add to your client config
Paste the URL into Claude Desktop, Claude Code, or Cursor settings. Each client has a slightly different config format — see the tabs below.
03
Authenticate on first use
The first time your client connects, Signa will prompt for your API key. Use the same key as your REST API key from Dashboard → Settings → API Keys.

Client configuration

// claude_desktop_config.json
// Claude Desktop → Settings → Developer → Edit Config
{
  "mcpServers": {
    "signa": {
      "url": "https://api.getsigna.ai/mcp/sse"
    }
  }
}
// Restart Claude Desktop. Authenticate on first use.

Prompt examples that work

📊"Is NVDA bullish or bearish — show me the key drivers"
🔍"What are the top 5 Tier 3 signals firing right now?"
📰"What's the news sentiment on AAPL this week?"
🌍"What's the current market regime and how does it affect signals?"
💼"Show me unusual options flow on META from the last 24 hours"
🏆"Which of the 40 agents has the highest Sharpe ratio this month?"

Available tools (8)

get_signalsEarly+
Returns nightly consensus signals filtered by tier, direction, and/or ticker. Includes drivers and risks.
get_fundamentalsEarly+
Valuation ratios, margin trends, and segment revenue breakdown.
get_newsAll plans
Recent news with pre-scored per-ticker sentiment. Multi-source: Benzinga, Yahoo, CNBC.
get_regimeAll plans
Current market regime classification and multiplier used to gate all signals.
screen_tickersEarly+
Run the screener with filters: Weinstein stage, RSI, sector, signal tier, market cap.
get_options_flowEarly+
Unusual Whales options flow — net call/put premium, institutional vs retail splits.
get_leaderboardAll plans
40-agent leaderboard with Sharpe ratios, win rates, and live accuracy metrics.
get_company_profileAll plans
Company name, sector, industry, description. Available on all plans.
03 — A2A PROTOCOL

Agent-to-Agent.
Attribution preserved.

The A2A layer is designed for AI pipelines. Your agent calls Signa, receives signals with confidence intervals and chain metadata, and passes them downstream — with provenance intact at every hop.

What makes A2A different from REST

The REST API returns signals formatted for humans. A2A returns signals formatted for agents — with additional fields that allow downstream agents to reason about signal provenance, not just signal value.

Every A2A response includes a watermark_id — a unique fingerprint per request that lets Signa detect if signal data is being redistributed without authorization. And a confidence_interval — a statistical range that downstream agents can use to assess signal reliability, not just the point estimate.

Most importantly: the origin_source field is immutable and must be preserved at every step of the chain. This is a legal requirement, not a preference.

Use cases

Multi-agent research pipeline
Research agent → Signa signals → Portfolio agent (sizing) → Risk agent (approval). Attribution preserved at every hop.
Automated morning briefing
Scheduler agent calls Signa at 6am → summarization agent distills → notification agent delivers. Each agent knows source.
Signal augmentation
Your proprietary model calls Signa as a "second opinion," blends outputs, and produces a combined recommendation with both sources attributed.
A2A REQUEST FLOW
Your Agent
──→
POST /v1/a2a/query
──→
Signa Scoring Engine
Downstream Agent 3
←──
Downstream Agent 2
←──
Signal + watermark_id + origin_source
origin_source: "Signa/FieldCrestVentures" — preserved at every hop — required by Terms of Use

Request & response

{  
  "agent_id": "agt_your-registered-id",   ← Required. Register in Dashboard → A2A Agents
  "agent_name": "ResearchAgent-v2",
  "query_type": "signals",
  "params": {
    "tickers": ["NVDA", "AAPL", "META"],
    "tier_min": 2,
    "include_drivers": true,
    "include_risks": true
  },
  "downstream_chain": ["PortfolioAgent-v1", "RiskAgent-v3"]   ← Optional. For audit trail.
}

Attribution rules

01
Preserve origin_source at every hop
Every data structure derived from a Signa signal must carry origin_source: "Signa/FieldCrestVentures". It must not be deleted, overwritten, or excluded at any layer of an A2A chain — even when your agent enriches or transforms the data.
Violation: Terms of Use §3.3 · Liquidated damages $250k/occurrence
02
No shadow modeling
You may not store, accumulate, or analyze patterns of Signa signals over time to train, calibrate, or improve any competing signal model. Even if origin_source is preserved, using signal patterns as training data violates the Terms of Use.
Violation: Signal Distillation — Exhibit A, Terms of Use v1.1
03
Register your agent before calling A2A
All A2A requests require a registered agent_id. Register in Dashboard → Settings → A2A Agents. Unregistered IDs are rejected with HTTP 403. This enables audit trails and rate limiting per agent, not just per account.
⚠️A2A protocol is available on Founding Member plan only. Enterprise or institutional use with custom rate limits and a Data Processing Agreement — contact legal@fieldcrestventures.com.

Start building on Signa.

Founding Member access includes REST API, MCP Server, and A2A Protocol — plus all 40 agent signals, Tier 3 alerts, and fundamental data.