v2.3.4 — MCP Server for Agent Ground Truth

Structured Ground Truth for AI Agents

AI agents working on your codebase need to understand project state. PulseTel is an MCP server that provides structured ground truth — CI, dependencies, coverage, issues, trends — via 12 standardized tools with a schema contract agents can rely on.

$ npx pulsetel-cli check

View on GitHub npm install

The Problem

AI agents working on codebases face fragmented health data:

CI status is in GitHub Actions
Security in npm audit
Coverage in a report
No persistent history
No trend analysis

Agents could DIY, but don't. It's complex and time-consuming.

The Solution

PulseTel is an MCP server that provides structured ground truth:

12 MCP tools — standardized schema contract
Telemetry collection — CI, deps, coverage, git, issues, PRs, Sentry
History persistence — track over time
Statistical analytics — trends, anomalies, confidence
Agent guidance — cross-signal correlation, reasoning prompts

MCP server provides ground truth. Agents interpret + act.

What PulseTel Provides

📊

Structured Telemetry

9 check types (CI, deps, coverage, git, health, issues, PRs, Sentry, deploy) — all normalized to a common schema.

📈

Trend Analysis

Direction, delta, velocity calculations. Know if coverage is degrading before it hits threshold.

🔍

Anomaly Detection

Statistical detection (2σ threshold) — not naive thresholds. Real anomalies, not noise.

🎯

Confidence Scoring

Multi-factor health aggregation (0-100). One number tells you project health.

🤖

Agent Guidance v2.3.4

_agent_guidance field with cross-signal correlations, observations, and investigation prompts.

⚡

Delta Mode

Token-efficient output showing only significant changes — 90% smaller than full check output.

MCP Server for Ground Truth NEW

PulseTel exposes 12 MCP tools that provide structured, schema-contracted ground truth about any project. Agents query — PulseTel answers with normalized data, trends, and cross-signal correlations they can act on.

{
  "results": [...],
  "_agent_guidance": {
    "observations": [
      "Coverage dropped 15.5% after dependency update",
      "1 new critical vulnerability introduced"
    ],
    "correlations": [{
      "checks": ["coverage", "deps"],
      "type": "causal_suspected",
      "description": "Coverage dropped after dependency changes",
      "investigation_prompt": "Check if dependency updates broke tests"
    }],
    "decision_tree": {
      "recommendation": "Address critical vulnerabilities first",
      "confidence": 0.95
    }
  }
}
            