Decision learning for AI agent actions. Evaluate, score, decide, and learn from outcomes.
Decision learning for AI agent actions. Evaluate, score, decide, and learn from outcomes.
Valid MCP server (2 strong, 1 medium validity signals). No known CVEs in dependencies. Package registry verified. Imported from the Official MCP Registry.
7 files analyzed · 1 issue found
Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.
This plugin requests these system permissions. Most are normal for its category.
Set these up before or after installing:
Environment variable: BIGHUB_API_KEY
Environment variable: BIGHUB_BEARER_TOKEN
Environment variable: BIGHUB_BASE_URL
Add this to your MCP configuration file:
{
"mcpServers": {
"io-bighub-mcp": {
"env": {
"BIGHUB_API_KEY": "your-bighub-api-key-here",
"BIGHUB_BASE_URL": "your-bighub-base-url-here",
"BIGHUB_BEARER_TOKEN": "your-bighub-bearer-token-here"
},
"args": [
"-y",
"@bighub/bighub-mcp"
],
"command": "npx"
}
}
}From the project's GitHub README.
Decision learning for AI agents.
BIGHUB is a decision layer that evaluates agent actions, returns structured recommendations, and learns from real outcomes over time.
Agent proposes action
↓
BIGHUB evaluates in context
↓
returns recommendation + confidence + rationale
↓
agent / runtime decides what to do
↓
real outcome is reported
↓
future recommendations improve over time
| Package | Language | Install | Description |
|---|---|---|---|
| bighub | Python | pip install bighub | Core SDK — evaluate actions, receive recommendations, report outcomes, retrieve learned signals. |
| bighub-openai | Python | pip install bighub-openai | OpenAI Responses API adapter — evaluate tool calls, surface recommendations, learn from outcomes. |
| @bighub/bighub-mcp | TypeScript | npm install @bighub/bighub-mcp | MCP server — evaluate agent actions from any MCP client. |
| bighub-anthropic | Python | — | Anthropic adapter — coming soon. |
| bighub-openai (JS) | TypeScript | — | OpenAI adapter for Node.js — coming soon. |
| Static guardrails | BIGHUB |
|---|---|
| Fixed allow / block logic | Structured recommendation in context |
| Rules only | Constraints + precedents + calibration + outcomes |
| One-off judgment | Trajectory- and history-aware evaluation |
| Prediction only | Prediction compared with real outcomes |
| Static thresholds | Recommendation quality improves over time |
pip install bighub
import os
from bighub import BighubClient
client = BighubClient(api_key=os.getenv("BIGHUB_API_KEY"))
# 1. Submit an action for evaluation
result = client.actions.submit(
action="update_price",
value=150.0,
domain="pricing",
actor="AI_AGENT_001",
)
# 2. Inspect the recommendation
print(result["recommendation"]) # proceed, proceed_with_caution, review_recommended, do_not_proceed
print(result["recommendation_confidence"]) # high, medium, low
print(result["risk_score"]) # 0.0 – 1.0
# 3. Act based on the recommendation
if result["recommendation"] in ("proceed", "proceed_with_caution"):
execute_action()
# 4. Report the real outcome
client.outcomes.report(
request_id=result["request_id"],
status="SUCCESS",
description="Price updated, no negative impact observed",
)
elif result["recommendation"] == "review_recommended":
request_human_review()
client.close()
Core loop: evaluate → recommend → act → report outcome → learn.
BIGHUB primarily returns:
recommendation — what to do (proceed, proceed_with_caution, review_recommended, do_not_proceed)recommendation_confidence — how confident (high, medium, low)risk_score — aggregated risk (0–1)enforcement_mode — how the recommendation is applied (advisory, review, enforced)decision_intelligence — rationale, evidence status, trajectory health, alternativesLegacy fields such as allowed, result, and reason may still appear for backward compatibility, but they are not the primary product surface.
BIGHUB evaluates actions not only in isolation, but also in the context of what happened before. As outcomes accumulate, similar sequences and prior decisions improve future recommendations.
For costly and multi-step workflows, trajectory-aware signals mean the same action may be judged differently depending on what happened earlier in the sequence.
Current Free BETA limits:
├── sdk/
│ └── python/
├── adapters/
│ ├── python/
│ │ ├── openai/
│ │ └── anthropic/
│ └── js/
│ └── openai/
├── servers/
│ └── mcp/
└── examples/
Be the first to review this server!
by Modelcontextprotocol · Developer Tools
Read, search, and manipulate Git repositories programmatically
by Toleno · Developer Tools
Toleno Network MCP Server — Manage your Toleno mining account with Claude AI using natural language.
by mcp-marketplace · Developer Tools
Create, build, and publish Python MCP servers to PyPI — conversationally.
by Microsoft · Content & Media
Convert files (PDF, Word, Excel, images, audio) to Markdown for LLM consumption
by mcp-marketplace · Developer Tools
Scaffold, build, and publish TypeScript MCP servers to npm — conversationally
by mcp-marketplace · Finance
Free stock data and market news for any MCP-compatible AI assistant.