symbiote-plugins

---

The Problem

Every AI coding session starts from scratch. The AI doesn't know your architecture, ignores your conventions, and repeats the same mistakes you corrected yesterday. Static rule files are manual and fragile. Every new conversation is a cold start.

Your corrections vanish. Your preferences reset. Your AI has amnesia.

Why Not Just CLAUDE.md?

CLAUDE.md is great for explicit instructions — but it's static text you maintain by hand. Symbiote is a living knowledge graph that updates itself every time your codebase changes.

	CLAUDE.md	Symbiote
Project structure	You write and maintain it	Auto-scanned, always current
Dependencies & impact	Not possible	Full graph — "what breaks if I change this?"
File connections	You document them manually	Every import, call, and edge mapped
Coding style	You write rules once	Learns from corrections, carries across projects
Stays current	Only if you remember to update it	Re-indexes on every change

They're complementary. Use CLAUDE.md for explicit instructions your AI should always follow. Use Symbiote for the deep project understanding no static file can provide.

---

Quick Start

As a Claude Code Plugin

/plugin marketplace add MohmmedAshraf/synapse
/plugin install symbiote@symbiote-plugins

No setup required — the plugin auto-installs the CLI, registers the MCP server, hooks, skills, and agents.

Manual Install (all editors)

npm install -g symbiote-cli
symbiote install

Auto-detects your installed editors (Claude Code, Cursor, Windsurf, Copilot, OpenCode), registers the MCP server globally for each, and sets up 9 Claude Code hooks + skills. Run it once.

Then in any project, open Claude Code and run:

/symbiote-init

This scans your codebase, extracts your coding preferences, project constraints, and architectural decisions — all in one step.

Symbiote initialized — scanned 142 files, recorded 18 DNA entries, 5 constraints, 3 decisions.

No manual config. No copy-pasting. After the first init, Symbiote's SessionStart hook auto-scans and boots the server on every new Claude Code session — zero cold start.

---

How It Works

Symbiote has two layers of intelligence that work together: Developer DNA (who you are) and the Project Brain (what your code is).

sequenceDiagram
    participant Brain as .brain/<br>Code Graph
    participant Sym as Symbiote
    participant AI as Claude Code
    participant DNA as ~/.symbiote/<br>Developer DNA

    Brain->>Sym: project structure, dependencies, health
    DNA->>Sym: coding rules, reasons, scoping

    loop Every interaction
        Sym->>AI: injects context (MCP + hooks)
        AI->>Sym: actions observed
        Sym->>Brain: re-indexes changed files
        Sym->>DNA: records corrections
    end

    Note over DNA: confidence evolves across sessions

Developer DNA — Your Coding Identity

Lives at ~/.symbiote/profiles/. Follows you across every project. Shareable.

When you correct your AI — "no, use early returns" — Symbiote captures it with the rule, the reason why, and which languages it applies to. Same correction across three sessions? Auto-promoted. Explicit instruction? Approved immediately at full confidence.

Your DNA is a single JSON profile — portable, exportable, switchable:

symbiote dna              # Active profile summary
symbiote dna export       # Export to .dna.json — share it anywhere
symbiote dna import <file|url>  # Load someone else's DNA
symbiote dna switch <name>      # Swap profiles instantly

Import a senior developer's DNA and your AI immediately codes in their style. Switch back to yours anytime. Your DNA is never lost.

Project Brain — Your Codebase's Nervous System