Best Brick AI Skills & MCP Servers

Hello-world brick — returns the message it receives. Serves as a smoke test for the FocusMCP pipeline.

Token budget management — estimate and optimize token usage for AI agents.

Knowledge graph export — generate HTML, Mermaid, GraphML, Cypher, Obsidian, and wiki formats.

Composite brick — orchestrates smartread, cache, compress, tokenbudget, and overview to deliver optimal context within a token budget.

Multi-source research — collect info from multiple files/sources, synthesize findings, validate consistency.

Shell command execution — run commands, background processes, kill, and compress output.

API route detection — scan projects for Express, Fastify, Next.js, SvelteKit route definitions.

Write files — create, overwrite, append.

Graph clustering — detect communities, explain groupings, identify architecture layers.

Pack files into a compressed context bundle — reduce token usage by extracting only signatures, maps, or full content.

Task dispatch queue — send tasks, manage queue, cancel, check status for multi-step workflows.

Sandboxed code execution — run JavaScript/TypeScript snippets in an isolated VM context.

AI team coordination — loads dispatch, parallel, debate, review, agent for multi-agent orchestration.

Inline, extract, and move refactoring — inline variables/functions, extract code to new functions, move between files.

Structured decision-making — define options, analyze tradeoffs, recommend, and record decisions.

File access heatmap — track read/write patterns, detect hot and cold files.

Track and report token savings — compare FocusMCP usage vs naive full-read, trend over time, ROI.

Semantic search using TF-IDF vector space — cosine similarity, intent matching, document similarity.

Task management for multi-agent workflows — create tasks, assign to agents, track status, mark complete.

Parallel execution coordination — run tasks concurrently, collect results, merge outputs, handle timeouts.

Typed SDK for building FocusMCP bricks — defineBrick() helper with full type safety for tools, manifests, and bus handlers.

Batch file reading — multiple files in one call, deduplication.

Validate data — JSON syntax, JSON Schema compliance, TypeScript type checking, lint rules.

Focus your AI agents on what matters. 68+ bricks, 1 MCP server, modular context — from 200k to 2k tokens. Works with Claude Code, Cursor, Codex.