deploysquad-recon-core 0.1.6

Build a structured AI knowledge graph in Obsidian through natural conversation

🛰️ recon — Obsidian × AI Knowledge Base

Your project's brain — built by AI, lived in Obsidian.

recon is a Claude Code MCP server that turns a natural conversation into a structured project knowledge graph — stored as Markdown in your Obsidian vault.

You describe your project. Claude infers goals, personas, modules, decisions, and features — then writes them as linked nodes. No forms. No templates.

The graph then feeds focused context back to any AI session via generate_context() — so Claude always knows what you're building, every session.

---

Install

/plugin marketplace add deploysquad-ai/recon
/plugin install recon@deploysquad-ai/recon

Then run /recon.setup, restart Claude Code, and run /recon.

---

The loop

💬 /recon       →   🧠 graph         →   ⚡ CONTEXT.md    →   🏗️ build
describe             10 node types        per feature           with context
project                                                              ↓
                ←─────────────── 🔄 /recon.add-feature ───────────┘
                                  keep graph current

---

Node types

Project → Goals → Personas → Constraints → Modules → Decisions → User Stories → Epics → Features → Versions

All stored as [[wikilinked]] Markdown in your Obsidian vault.

---

Python API

from deploysquad_recon_core import (
    create_node, get_node, list_nodes, update_node,
    resolve_links, build_index, generate_context,
)

# Author a node
path = create_node("feature", {
    "name": "Task Board",
    "description": "Kanban board for task management",
    "implements": ["[[User Story - Create Task]]"],
    "actors": ["[[Persona - Manager]]"],
    "belongs_to": "[[Module - Dashboard]]",
    "status": "active",
}, project_dir)

# Generate CONTEXT.md for any AI session
context = generate_context("Task Board", project_dir)

---

Semantic linking (optional)

recon can embed nodes and find semantically similar ones using the Gemini API.

Claude Code plugin users: /recon.setup prompts for your Gemini API key and writes it to the MCP server config automatically.

Python library users:

pip install "deploysquad-recon-core[embed]"

Set GEMINI_API_KEY in your environment, then:

from deploysquad_recon_core import embed_nodes, find_similar

embed_nodes(project_dir)
similar = find_similar(node_path, project_dir)

---

Links

• GitHub
• Documentation