Apache-2.0 · Deterministic MCP Server · Zero LLM Calls

Your AI Coding
Advisory Squad

A deterministic MCP server that transforms your AI coding assistant into a multi-agent advisory team. Classify tasks, score risk, assemble specialist reviewers, implement with human gates, and build persistent project memory — all with zero LLM calls in the server.

13
Skills
20+
MCP Tools
11
Specialist Agents
0
Server LLM Calls

What Makes It Different

Most AI coding tools are black boxes. squad-mcp gives you a transparent, auditable, and configurable advisory layer between you and your AI assistant.

Deterministic Server

Zero LLM calls in the server. Every tool is a pure function — classification, risk scoring, agent selection, and consolidation are all heuristic and auditable.

🛡

Human Gates

Two explicit gates in every implementation run: plan approval and Blocker halt. The squad never writes code you didn't sign off on. Ever.

🧠

Risk-Adaptive Depth

Auto-scales from quick (2 agents, sub-30s) to deep (5+ agents, architect + security forced). Override with --quick / --deep flags.

💡

Persistent Memory

.squad/learnings.jsonl records every accept/reject decision. The squad stops re-raising findings your team already resolved. Decisions are versioned in git.

📋

PRD Decomposition

Feed a PRD, get atomic tasks with dependencies, scope globs, and agent hints. Each task narrows the squad's focus — no more re-analyzing the whole repo.

📈

Weighted Scorecard

Every review produces a weighted rubric scorecard (0–100) per dimension. Configurable thresholds per repo. APPROVED, CHANGES_REQUIRED, or REJECTED.

From Prompt to Reviewed Code

A single /squad:implement command threads through classification, planning, advisory, and consolidation — with human approval at every critical point.

1

Classify & Score Risk

The server analyzes your prompt + changed files, classifies the work type (Feature, Bug Fix, Security, etc.), and computes a Low/Medium/High risk score from boolean signals (auth, money, migration, file count).

2

Select Squad & Pick Depth

Based on the work type and risk, specialist agents are selected. Depth auto-resolves: quick for low-risk, deep for high-risk/security, normal otherwise. Each agent gets only the files in their domain.

3

Draft Plan & Gate 1

A plan is drafted and sent to the tech-lead-planner for review (skipped in quick mode). The skill stops and asks you to approve. Reply "go" to proceed; anything else cancels.

4

Parallel Advisory Review

Every selected agent reviews in parallel, emitting findings + a Score: NN/100. Architect, DBA, Developer, QA, Security, Reviewer — each sees only their slice.

5

Consolidate & Gate 2

The tech-lead-consolidator produces a verdict (APPROVED / CHANGES_REQUIRED / REJECTED) with a weighted scorecard. If any Blocker is found, the squad halts and asks you.

6

Implement & Record

If approved, the implementer writes the code. Never commits or pushes — that's your call. The run is recorded to .squad/runs.jsonl for observability.

13 Skills at Your Fingertips

Every command is a standalone skill. Chain them with /squad:pipeline for a full cradle-to-grave workflow.

/squad:implement <task>
Full implementation workflow: classify, plan, gate, advisory squad, consolidate, implement. Two human gates.
implement
/squad:review [target]
Review an existing diff, branch, or PR. Same advisory squad, no implementation. Posts verdict + scorecard.
review
/squad:question <question>
Fast read-only code Q&A. Spawns code-explorer, answers with file:line citations. No plan, no gates.
read-only
/squad:debug <issue>
Read-only bug investigation. Emits ranked hypotheses with file:line evidence and verification steps.
read-only
/squad:tasks <prd>
Decompose a PRD into atomic tasks with dependencies, scope, and agent hints. Confirmation before saving.
pipeline
/squad:next · /squad:task <N>
Pick the next ready task or work on a specific one. Squad narrows scope to the task's glob.
pipeline
/squad:grillme <plan>
Socratic plan validation. Grills your plan against CONTEXT.md and ADRs, resolving terms inline.
pipeline
/squad:pipeline <feature>
Cradle-to-grave: brainstorm → grillme → tasks → next → implement → review. Stateful advisor.
pipeline
/brainstorm <topic>
Exploratory Q&A with web research + specialist perspectives. Options matrix with cited sources. No code.
utility
/squad:stats
Observability dashboard. Verdict mix, score buckets, sparkline trend, per-agent wall-clock and token breakdown.
utility
/commit-suggest
Generate Conventional Commits messages for staged changes. Read-only, never mutates git.
utility
/squad:enable-journaling
Opt-in auto-journaling via PostToolUse hooks. Captures every run to .squad/runs.jsonl.
utility

See It in Action

Every example is a single line you type. The squad sizes itself from the prompt + changed files. You only reach for a flag to override.

Low-Risk Feature — Auto-Detected Quick quick
/squad:implement add a /health endpoint that returns {"status":"ok"}
work_type: Feature · risk: Low · mode: quick (auto) · agents: [developer, qa]
Planner skipped, 2-agent advisory, sub-30s feedback. Stops at Gate 1 for your approval.
Auth Refactor — Auto-Detected Deep deep
/squad:implement refactor src/auth/jwt-validator to rotate signing keys
work_type: Security · risk: High · mode: deep (auto) · agents: [architect, security, developer, qa, reviewer]
touches_auth fires → deep. Planner + consolidator personas run, reject-loop ceiling raised to 3.
Safety Override — Forcing Quick on Risky Diff quick + safety
/squad:implement --quick patch the refund amount rounding in src/billing/ledger.ts
mode: quick (user) · mode_warning set — --quick is honoured but security is force-included as one of the 2 agents because touches_money fired.
Fast Code Q&A — No Plan, No Gates read-only
/squad:question where is the rubric weighted score computed?
Spawns code-explorer, answers with file:line citations. Sub-second on --quick. No plan, no gates, no implementation.

Up and Running in 60 Seconds

Four ways to install. Pick your host and go.

Claude Code
/plugin marketplace add ggemba/squad-mcp
/plugin install squad@gempack
The plugin bundles the MCP server, slash commands, and skills. After install, restart Claude Code to pick up the new commands and the squad MCP server.
Terminal
npx -y @gempack/squad-mcp
The package exposes the squad-mcp binary and works with any MCP-capable client via stdio.
claude_desktop_config.json
{
  "mcpServers": {
    "squad": {
      "command": "npx",
      "args": ["-y", "@gempack/squad-mcp"]
    }
  }
}
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
.cursor/mcp.json
{
  "mcpServers": {
    "squad": {
      "command": "npx",
      "args": ["-y", "@gempack/squad-mcp"]
    }
  }
}
Workspace-scoped .cursor/mcp.json or global Cursor settings. Same config shape works for Warp too.

Deterministic by Design

The server makes no LLM calls. The host LLM does all reasoning; the server hands it building blocks.

Host LLM
Claude Code / Cursor / Warp
squad-mcp Server
Tools · Prompts · Resources
Specialist Agents
Host orchestrates → Server provides deterministic tools → Agents review in parallel → Scorecard returned

11 Specialist Agents

Each agent has a domain-specific system prompt and reviews only the files in their scope.

product-owner
planner
consolidator
architect
security
dba
developer
reviewer
qa
code-explorer
debugger
SQUAD RUBRIC — WEIGHTED 82 / 100 APPROVED
Application Code
82 ×18%
Testing & QA
78 ×14%
Architecture
85 ×20%
Security
90 ×25%
Database
72 ×12%
Business Logic
80 ×11%