Simone MCP + PCPM — Mandatory Agent Intelligence Layer

MANDATE (PRIORITY -9.5): Every OpenSIN agent, every worker, every coder must activate Simone MCP and load PCPM context before any code navigation, symbol editing, structural change, or implementation task begins. This is not optional.

---

Why Simone MCP?

AI coding agents that operate without semantic code understanding are blind. They guess file locations, copy-paste symbols without understanding call graphs, and create regressions by editing in the dark.

Simone MCP is the OpenSIN semantic intelligence server that gives every agent:

• Symbol navigation — find exactly where a function, class, or variable is defined across the entire workspace
• Reference discovery — know every call site before you rename or refactor
• Structural editing — replace symbol bodies via LSP, not fragile regex hacks
• Project overview — language-aware codebase map for any directory
• Hybrid memory — Qdrant + Neo4j vector/graph store for cross-session knowledge
• A2A + MCP surface — streamable HTTP MCP at /mcp, A2A JSON-RPC at /a2a/v1, health at /health, metrics at /metrics

---

Why PCPM?

Without persistent context, every agent session starts from scratch. The Global-Brain DPMA v4 / PCPM system ensures:

• No amnesia between sessions — architectural decisions, bug history, and coding conventions survive restarts
• Hive mind synchronization — all agents share the same knowledge base
• Plan continuity — active tasks are resumed, not restarted
• Evidence-first audit — every decision is backed by stored context, not hallucination

---

Activation Protocol (MANDATORY before any coding task)

Step 1 — Check PCPM / Global-Brain coupling

# In every project root, before any other work:
node /Users/jeremy/dev/global-brain/src/cli.js setup-hooks \
  --project $(basename "$PWD") \
  --project-root "$PWD" \
  --agents-directive

If .pcpm/ or .opencode/opencode.json with hooks already exists → skip, already coupled.

Step 2 — Verify Simone MCP is reachable

# Health check (local):
curl -s http://localhost:8765/health | python3 -m json.tool

# Health check (HF Space — production):
curl -s https://openjerro-simone-mcp.hf.space/health | python3 -m json.tool

Expected response:

{
  "status": "ok",
  "service": "simone-mcp",
  "version": "1.0.0",
  "transports": ["stdio", "streamable-http"],
  "tools": 10,
  "capabilities": ["symbol_navigation", "references", "structural_edit", "memory"]
}

Step 3 — Use Simone MCP for ALL code operations

Task	Correct approach	Forbidden approach
Find where UserService is defined	find_symbol("UserService") via Simone MCP	Grepping blindly through files
Rename getUser everywhere	replace_symbol_body + LSP rename via Simone MCP	sed/awk across the repo
Understand the codebase before editing	get_project_overview() via Simone MCP	Reading random files and guessing
Find all callers of processPayment	find_references("processPayment") via Simone MCP	Manual grep
Insert a new method into a class	insert_after_symbol via Simone MCP	Raw file append

---

OpenCode Integration

Add Simone MCP to your project's OpenCode config (.opencode/opencode.json):

{
  "mcp": {
    "simone-mcp": {
      "type": "streamable-http",
      "url": "https://openjerro-simone-mcp.hf.space/mcp",
      "description": "Semantic code intelligence — symbol navigation, references, structural editing, hybrid memory"
    }
  }
}

For local development, replace the URL with http://localhost:8765/mcp.

---

Simone MCP Tool Reference

Tool	Description
find_symbol	Find all definitions of a symbol by name path pattern
find_references	Find all usages/call sites of a symbol
replace_symbol_body	Replace a symbol's full body (LSP-grade)
insert_after_symbol	Insert code immediately after a symbol definition
get_project_overview	Get a language-aware overview of the codebase structure
execute_simone_action	Run a named high-level action (refactor, validate, audit)
process_lsp_task	Submit an LSP task for async processing
memory_store	Persist a knowledge item to hybrid vector/graph memory
memory_query	Retrieve relevant knowledge from hybrid memory
health_check	Verify Simone MCP is alive and responsive

---

Repository

• GitHub: https://github.com/OpenSIN-AI/Simone-MCP
• HF Space: https://huggingface.co/spaces/openjerro/simone-mcp
• MCP endpoint: https://openjerro-simone-mcp.hf.space/mcp
• A2A endpoint: https://openjerro-simone-mcp.hf.space/a2a/v1
• Dashboard: https://openjerro-simone-mcp.hf.space/dashboard

---

Enforcement

This mandate is enforced at multiple levels:

1. Global AGENTS.md (~/.config/opencode/AGENTS.md) — Priority -9.5 rule
2. upgraded-opencode-stack installer — checks for Simone MCP coupling on install
3. MANDATE-0.34.md — canonical blueprint mandate file in upgraded-opencode-stack
4. worker-code.md — explicit <orchestration_mandate> block in every worker instruction
5. Template-A2A-SIN-Agent AGENTS.md — every new agent inherits this requirement

Violation = architectural failure. If an agent edits code without Simone MCP + PCPM active, its changes are considered unverified and must be reviewed before merge.

---

Quick Checklist

Before starting any coding task:

• [ ] PCPM / Global-Brain hooks are installed in this project (.opencode/opencode.json exists with beforeRun/afterRun hooks)
• [ ] Simone MCP health check returns "status": "ok"
• [ ] OpenCode MCP config includes simone-mcp entry
• [ ] No symbol edits are made via raw file search/replace without first running find_symbol or find_references

---

Last updated: 2026-04-12 Status: ACTIVE & MANDATORYMaintainer: sin-zeus Mandate reference: MANDATE-0.34.md