chisel_

Your codebase grows faster than your memory.
Chisel keeps you the architect.

Structured specs and docs your LLM can actually use. Write once, query from anywhere. The project brain for solo founders and small teams building with AI agents.

We are currently in public alpha. Sign up for release notes.

 $ curl -sL https://install.chisel.build
                        | sh

Public Alpha

chisel — demo

q:quit/:searchm:machine?:helphjkl navigation

The context problem

Solo founders and small teams using LLM agents ship fast — faster than any human can track. Six months in, the codebase has outgrown your ability to hold it in your head. Every new agent session starts from scratch. Decisions are made twice. Architecture drifts from intent.

Chisel is the infrastructure that preserves your reasoning. Specs capture intent, constraints, and decisions. Docs preserve permanent knowledge. Both are structured so your LLM tools can read them as context when they need it.

Available Now

New

Chisel Specs

Lifecycle-aware specs for active work. Draft → ready → in-progress → shipped. The highest-value context artifact for your LLM mid-implementation.

Stable

Chisel Docs

Permanent knowledge base with 3-pane TUI explorer, SQLite full-text search, and structured machine mode output.

Chisel Specs

The artifact that lives between an idea and a shipped feature. Captures intent, constraints, alternatives, and implementation notes. Structured for both humans and LLM agents.

Lifecycle Machine Mode Templates

Chisel Docs

A knowledge base that understands structure. Markdown-first. Local SQLite index. AI-ready schemas. Your permanent project memory.

Markdown Local Index RAG Ready

Without Chisel vs. With Chisel

	Without Chisel	With Chisel
Decision memory	Lost in Slack threads and closed PRs	Preserved in versioned specs alongside code
Agent context	Every session starts from scratch	One command generates structured context
Spec lifecycle	Google Docs that go stale after shipping	Draft → ready → in-progress → shipped → archived
Knowledge retrieval	Manual copy-paste from wikis	Full-text search with machine mode output

Decision Memory

Without

Lost in Slack threads and closed PRs

With Chisel

Preserved in versioned specs alongside code

Agent Context

Without

Every session starts from scratch

With Chisel

One command generates structured context

Spec Lifecycle

Without

Google Docs that go stale after shipping

With Chisel

Draft → ready → in-progress → shipped

Knowledge Retrieval

Without

Manual copy-paste from wikis

With Chisel

Full-text search with machine mode output

Built for humans. Structured for AI.

Every Chisel tool supports a mode optimized for LLM context windows and parsing. Works with Claude Code, Cursor, Aider, and any shell-capable agent.

BYOLLM: Chisel structures your knowledge. You bring the model.

Chisel Docs

Interactive TUI for Developers

Full terminal UI with colors, borders, and keyboard shortcuts (j/k) for efficient human navigation.

e:edit n:new Tab:switch hjkl navigation

The Workflow

# Write a spec for your next feature

$ chisel spec new "user authentication"

# Implement with your LLM agent

$ chisel context create "auth" | claude -p "implement this"

# Mark shipped when done

$ chisel spec status user-auth shipped

The spec now reflects what actually shipped. Six months later, you or your agent can find it.

01. Architecture

Written in Rust for near-zero overhead. TUI response times under 10ms on massive datasets. No Node.js or Python runtime required.

02. Storage

Local-first Markdown & YAML files. An internal SQLite FTS5 index provides instant full-text search.

03. License

Chisel is FSL-1.1. Free for non-competing use and converts to Apache 2.0 after 2 years.

04. Extensible

Built to be extended. Use the Machine Mode YAML output to pipe Chisel into scripts or AI agents.

Start staying the architect.

Chisel is in active development. Try the alpha today and keep your project's reasoning intact as it scales.

View Documentation View on GitHub

v0.2.1-alpha • FSL-1.1-Apache-2.0 Licensed