Aegus

mcp toolkit · the institute as an api

Call the Institute from your own agents.

The same grading framework that powers PATH trading, voice calls, content, and reports — exposed as MCP tools. Call from Claude, ChatGPT, or your own agent infrastructure. One unified reasoning engine, four callable tools.

The four tools.

grade_token

POST /grade

Grade a Solana token for paper or live entry.

Request body
{
  "entity_type": "trade",
  "entity_id": "PNUTpump...",
  "payload": {
    "smart_wallet_count": 5,
    "social_heat": 82,
    "creator_credibility": 0.9,
    "rug_score": 10,
    "momentum": 12.3,
    "market_cap_usd": 350000
  }
}
Sample response
{
  "grade": "A",
  "score": 92.9,
  "thesis": "5 smart wallets entered within a narrative-heat window; rug profile clean, momentum up.",
  "strength_tags": [
    "smart_wallet_strong",
    "narrative_hot",
    "clean_rug_profile"
  ],
  "confidence": 0.95
}

grade_call

POST /grade

Grade a voice-agent call on ICP fit, urgency, objection, and pattern match.

Request body
{
  "entity_type": "voice_call",
  "entity_id": "call_20260415_acme",
  "payload": {
    "icp_fit": 0.88,
    "urgency": 0.75,
    "objection_type": "none",
    "past_pattern_wr": 0.52,
    "budget_signal": 0.85,
    "prospect": "Acme Dental"
  }
}
Sample response
{
  "grade": "A",
  "score": 75.2,
  "thesis": "Prospect well fits ICP with high urgency; pattern-library says 52% of similar calls close.",
  "strength_tags": [
    "strong_icp_match",
    "urgent_timeline",
    "budget_confirmed"
  ],
  "confidence": 0.9
}

grade_content

POST /grade

Grade a piece of content — hook + niche + thumbnail + first-3s opening.

Request body
{
  "entity_type": "content",
  "entity_id": "clip_crypto_2026_q2_001",
  "payload": {
    "hook_pattern_wr": 0.62,
    "niche_match": 0.92,
    "thumbnail_score": 0.85,
    "first_3s_score": 0.8,
    "length_fit": 0.9,
    "niche": "crypto"
  }
}
Sample response
{
  "grade": "A",
  "score": 79,
  "thesis": "Hook pattern lands at 62% CTR in this niche; thumbnail strong; first-3s gripping.",
  "strength_tags": [
    "proven_hook_pattern",
    "niche_locked",
    "thumbnail_strong"
  ],
  "confidence": 0.89
}

grade_report

POST /grade

Grade a report on source density, confidence, agreement, and citation coverage.

Request body
{
  "entity_type": "report",
  "entity_id": "report_q1_crypto_memes",
  "payload": {
    "source_count": 12,
    "avg_confidence": 0.82,
    "source_agreement": 0.85,
    "insight_count": 9,
    "citation_coverage": 0.95,
    "contrarian_signals": 1
  }
}
Sample response
{
  "grade": "A",
  "score": 91.5,
  "thesis": "Report draws on 12 sources with 82% average confidence; sources align; 95% of claims carry citations.",
  "strength_tags": [
    "well_sourced",
    "high_confidence",
    "fully_cited"
  ],
  "confidence": 0.92
}

One endpoint, many callers.

All four tools are thin wrappers over a single endpoint. Same grading code path that PATH, the voice agent, Aegus Content, and Aegus Reports call internally. Routing is by entity_type; every response shape is identical.

# Any tool, same endpoint:
curl -X POST https://api.aegus.digital/institute/grade \
     -H 'Authorization: Bearer ${AEGUS_KEY}' \
     -H 'Content-Type: application/json' \
     -d '{
       "entity_type": "trade",
       "entity_id": "PNUTpump",
       "payload": { "smart_wallet_count": 5, "social_heat": 82, ... }
     }'

MCP manifest (for Claude / ChatGPT / agent frameworks).

{
  "mcpServers": {
    "aegus-institute": {
      "url": "https://mcp.aegus.digital",
      "tools": ["grade_token", "grade_call", "grade_content", "grade_report"],
      "description": "Institute grading framework — graded A/B/C with thesis + evidence chain"
    }
  }
}

Access.

internal

Already live inside Aegus

PATH, Front Desk, Capture, Clip, and Reports all call these tools internally today. Every verdict you see on /trading, /voice-agent, /content, and /reports comes from the same endpoint.

public

Public MCP — gated rollout

The public-facing MCP endpoint is in gated rollout. Early partners get credentialed access, a per-call quota, and priority on new grader domains. Request access →

Paired with the rest of the stack.

MCP Toolkit is the Institute exposed — same reasoning framework, same pattern library, same grading as every other Aegus service. See the full diagram.