AI Boost MCP server surfaces personal coding patterns across sessions
AI Boost is a personal MCP server that stores coding patterns and conventions as "boosters" — indexed by keywords and embeddings — and automatically surfaces them at the start of relevant agent tasks.
Score breakdown
Evaluate AI Boost as a way to stop re-explaining project conventions to coding agents on every session — the auto-suggest behavior before task start is the key UX question the author is seeking feedback on.
- 01Built by npiano to solve the problem of LLM agents starting every session without knowledge of personal patterns and conventions.
- 02Users save content as 'boosters' — currently supporting text files and public GitHub repos; private GitHub repo support is in progress.
- 03Boosters are indexed using both keywords and embeddings for automatic, context-aware retrieval.
AI Boost, posted to Hacker News by author npiano, addresses a common frustration with LLM-based coding agents: every new session starts blank, forcing developers to re-explain their patterns and conventions from scratch. The tool acts as a personal library backed by an MCP server, where users save content — currently text files or public GitHub repos — as named "boosters." Each booster is indexed with both keywords and embeddings so the agent can automatically surface relevant ones before a task begins.
The author cites auth flows and Terraform patterns for AWS as personal examples of the kind of reusable knowledge the tool targets.
The author cites auth flows and Terraform patterns for AWS as personal examples of the kind of reusable knowledge the tool targets. Existing approaches like rules, skills, and memory were found insufficient — rules and skills are hard to synchronize across multiple agents, projects, and machines, while memory is too noisy due to its unstructured nature. Boosters are private by default and scoped to the user's account. A community marketplace is in development that would allow users to publish boosters and earn credits each time one is injected, though this feature is not yet complete. The MCP endpoint is `https://mcp.ai-boost.io/mcp`, and the tool is intended to work with any MCP-compatible client including Cursor and Claude Code.
Key facts
- 01Built by npiano to solve the problem of LLM agents starting every session without knowledge of personal patterns and conventions.
- 02Users save content as 'boosters' — currently supporting text files and public GitHub repos; private GitHub repo support is in progress.
- 03Boosters are indexed using both keywords and embeddings for automatic, context-aware retrieval.
- 04Boosters are private by default and accessible only to the owner's account.
- 05A community marketplace is planned where users can publish boosters and earn credits per injection, but is not yet fully available.
- 06The MCP server endpoint is `https://mcp.ai-boost.io/mcp` and is designed to work with Cursor, Claude Code, and any MCP-compatible client.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 8, 2026 · 15:36 UTC. How this works →