Primary Logic Investment Intelligence

Use the Primary Logic MCP tools to retrieve real-time, continuously refreshed investment context with LLM-ranked relevance and impact signals for public and private companies.

Setup

Claude Code / Cowork (recommended)

claude plugins add PrimaryLogic/agent-skills

Codex

codex mcp add primary-logic --url https://primarylogic--pulse-backend-external-api-app.modal.run/mcp

Other MCP-capable agents

Add to your agent’s MCP config:

• URL: https://primarylogic--pulse-backend-external-api-app.modal.run/mcp
• Auth: OAuth (automatic) or Bearer token with API key

npx (universal)

npx skills add PrimaryLogic/agent-skills

Activation Cues

Activate this skill when the user asks for any of:

• ticker-specific bullish or bearish evidence
• recent catalysts or risk signals
• per-content relevance or impact details by ticker
• source coverage or content visibility checks
• usage diagnostics

What This Data Represents

• Source-normalized investment context: top podcasts, articles/news, X/Twitter, Kalshi, Polymarket, earnings calls, filings, and other monitored channels normalized into one feed
• LLM-heavy signal extraction: per-ticker relevance and impact scores attached to each content item to prioritize material developments
• Public + private company coverage: source visibility and ticker coverage for the requesting organization

Available MCP Tools

Use the primary-logic connector tools:

Hard Rules

• Never fabricate data; all claims must map to tool responses.
• Data visibility is org-scoped; if records are missing, org source visibility may be the cause.
• If a tool call fails, report the error and suggest a concrete next step.
• Use absolute timestamps in outputs when the user asks about recent windows.
• Do not claim market prices, positions, or execution events unless explicitly present in the data.
• Do not present outputs as guaranteed returns or autonomous execution instructions.
• Frame outputs as context for decision support and user-approved actions.

Input Contract

Interpret each user request into this query plan:

1. objective: thesis_support, counter_thesis, catalyst_scan, sentiment_shift, coverage_check
2. scope:
   • tickers: list of uppercase ticker symbols
   • time window: since and until in ISO datetime format
   • source_types: optional list
3. signal filters:
   • min_relevance: 0..1
   • min_abs_impact: 0..10
   • sentiment: positive | negative | neutral
4. retrieval:
   • limit (default 50)
   • sort mode: date | abs_impact | relevance

If ticker or time window is missing for an investment query, ask one concise clarification.

Decision Workflow

1. Validate connectivity with health_check.
2. Use search_content for broad discovery pulls.
3. Use get_ticker_content for signal-ranked ticker analysis.
4. Use get_content_ticker_signals for per-item attribution detail.
5. Use get_ticker_detail with include_signal_stats=true for summary context.
6. Use list_sources when visibility or source scope is ambiguous.
7. Use get_usage for rate-limit troubleshooting.

Output Contract

Return structured investment output with:

1. key_findings: 3 to 7 concise bullets
2. thesis_view: one short paragraph
3. supporting_evidence: list of {content_id, ticker, impact_score, relevance_score}
4. contrary_evidence: same schema
5. catalysts: list
6. risks: list

If results are empty, return “no qualifying records” and suggest which filter to relax first.

Query Defaults

• Default content limit: 50 unless user asks otherwise.
• Apply ticker filters whenever the user names tickers.
• For signal-heavy tasks, start with min_relevance >= 0.6 and min_abs_impact >= 5.