Skip to main content
Glama
Sign In
Agent Shadow Brain

Agent Shadow Brain

by ihtishaam-dev
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Local

What you can do

Workflow

What you can inspect or change

Review project memory

Search saved decisions and lessons, inspect recorded sources, and record reviews or typed relationships in the memory graph.

Build task context

Assemble a project-scoped pack within a size budget. See why each memory was included or skipped before handing the pack to an agent.

Follow live operations

Inspect measured MCP calls, explicitly reported task phases, and the event timeline. Pause updates or reduce motion.

Give agents focused workflows

Install six project skills for workflow selection, implementation, debugging, review, research, and release.

Check integrations

See supported client capabilities, project skill-file presence, storage health, and a copyable diagnostic report.

The workspace connects these views. Open a memory from a context pack or graph edge to see the saved content, its recorded references, and the review history. You can control context assembly, cross-project context, session briefings, live updates, and activity recording separately. Screenshot gallery · Workspace guide

Concept diagram of coding agents sharing project memories and receiving scoped context packs

Related MCP server: junto-memory

Quickstart

Use Node.js 22.12 or later in the 22.x line, 24.x, or 26 and later. Run the dashboard command from your project directory.

npm install -g @theihtisham/agent-shadow-brain@7.1.3
shadow-brain --version
shadow-brain dash . --no-analysis

The dashboard prints its local address. Open Memories to inspect saved records, Context to assemble a pack, or Live to follow recorded activity. The CLI also assembles a pack from the current project.

shadow-brain context-pack "import retry behavior" --project . --token-budget 1200 --json

Pin 7.1.3 for this release because npm’s latest tag and semantic version order can differ. The quickstart also includes the verified GitHub tarball and a first-memory example. Full quickstart

Connect a client

Print the standalone MCP launch configuration, then add it to your client’s MCP settings.

shadow-brain clients config

The launch command is shadow-brain mcp --stdio. It serves memory, context, and workflow tools locally without loading model-provider configuration. Use context_pack for curated project context and activity_report to explicitly record host activity. The live journal keeps the latest 500 events with bounded labels and timings. MCP and activity guide

Client

Skill target

Project skill directory

Codex

codex

.agents/skills/

Claude Code

claude-code

.claude/skills/

OpenCode

opencode

.opencode/skills/

DeepSeek Harness

deepseek-harness

.dsh/skills/

Other compatible clients

portable

.agents/skills/

Preview the skill files and apply the installation for your chosen target.

shadow-brain skills list
shadow-brain skills install . --target codex
shadow-brain skills install . --target codex --apply

Skill discovery and MCP connections depend on the client’s settings. DeepSeek Harness needs its filesystem skill and MCP client plugins enabled. The dashboard’s Integrations view checks managed project files. Client setup guide

Documentation

Guide

Covers

Start here

Installation, first memory, dashboard, and MCP setup

Workspace

Reviews, relationships, context limits, activity journal, and local API

Integrations

MCP setup and project skill layouts by client

Workflow skills

Installation layouts, previews, and the six bundled skills

Showcase

Annotated gallery of the workspace and live portal

Troubleshooting

Startup, memory, context, and integration checks

7.1.3 release notes

Upgrade steps, curation changes, and live operations

Download statistics

npm download counts and their reporting window

Documentation index · Changelog · Report an issue

Package details

Item

7.1.3 release

Package

@theihtisham/agent-shadow-brain

CLI

shadow-brain or asb

Interfaces

Local web workspace, CLI, MCP stdio, and JavaScript/TypeScript library

Node.js

^22.12.0 || ^24.0.0 || >=26.0.0

Tarball

3,443,588 bytes packed, about 8.7 MB unpacked, 729 files

License

MIT

Verified tarball · Package verification report

Maintainers

@ihtishaam-dev · @theihtishaam

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    This server cannot be deployed

    Claude Code, inside the Godot 4 editorFlockbay puts an AI agent in the real Godot 4 editor: it writes GDScript, runs your game and fixes what breaks. Free, or bring your own Claude Code or Codex.flockbay.comAd

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Resources

    Related MCP Connectors

    Related MCP Servers

    • LocalNest MCP

      wmt-mobile
      A
      license
      B
      quality
      D
      maintenance
      A local-first MCP server that provides AI agents with safe codebase access through file discovery, hybrid lexical-semantic search, and project introspection. It features durable local memory and semantic indexing while keeping all data and processing entirely on your local machine.
      74
      87 npm
      6
      MIT

    • knitbrain

      PDgit12
      A
      license
      A
      quality
      C
      maintenance
      Local-first MCP server that gives any AI coding agent per-project memory, workflow intelligence, and always-on, lossless token & context optimization.
      37
      65 npm
      6
      MIT