A fresh session starts blank.
Every "where were we?" is re-derived from scratch, or re-explained by you, for the tenth time.
// durable working memory for coding agentsYour coding agent is brilliant,
Your coding agent is brilliant,
and amnesiac.
Clauderizer gives it a memory that survives: plans, decisions, and a dependency graph as plain, Git-diffable Markdown it reads and updates through real tool calls.
- phases
- decisions
- invariants
- subsystems
What a session feels like
Your agent starts oriented, not blank.
Before you type a word, the SessionStart hook has already told the agent exactly where things stand. No prescribed reading order, no re-briefing. This is the digest for this very site.
you open a fresh session, nothing typed yet
[Clauderizer] gameplan marketing-site-launch · phase 4 / 10
baseline 0 tests · 10 lessons · 0 pending cascades
next: cz_next_phase_context → cz_preflight
↑ injected automatically: no prescribed reading order, no re-briefing
▸ I see where we are: phase 4 of 10. Preflight passed. Building the interactive demos now.
- 0
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
Every session picks up exactly where you left off.
That's how you fly. ✦
The failure mode
Long-running AI work hits two walls.
A long session fills its context.
Mid-task the window overflows and the agent forgets its own earlier decisions, then confidently contradicts them an hour later.
The usual answer is conventions: "keep a decisions log," "write a handoff." But conventions rot, because nothing executes them. Clauderizer keeps the proven model and makes it a system, not a hope.
Same agent, same repo
One of them remembers.
- You re-explain the architecture. Again.
- A decision from an hour ago, contradicted.
- “Wait… did we already build this?”
- The plan drifts from reality, silently.
every session starts from zero
- The digest greets the agent at session start.
- Decisions & invariants surface, not re-derived.
- “Phase 4 of 10, preflight passed, building.”
- The plan is the live state.
every session is momentum →
How it works
Markdown is the source of truth.
The graph index is a disposable cache, rebuilt from the Markdown on demand. If they ever disagree, Markdown wins. No database, no lock-in. Your memory is just readable files in docs/.
- 01Markdown in docs/
Plans, decisions, invariants, a dependency graph, all diffable, reviewable, and shipped with the code.
- 02The engine
Reads and writes the Markdown, rebuilds the graph cache, and runs preflight and cascade against the real repository.
- 03MCP tools + hooks
Self-describing tool calls and lifecycle hooks inject status into context, automatically.
- 04Your agent
Oriented before you type a word. No prescribed reading order, no re-briefing, no drift.
cz_next_phase_contextthe full bundle for this phasecz_preflightactually runs your tests / build… do the work …write code, honour the invariantscz_add_decision · cz_add_invariantstructured, graph-aware recordscz_cascadewalk the graph, flag dependentscz_write_handoffassemble the next cumulative handoff
Why it holds
A system, not a hope.
Prose conventions rot because nothing executes them. Clauderizer keeps the proven model and makes it real tool calls the agent can't forget to make.
Discoverable
Self-describing MCP tools + lifecycle hooks inject status into context automatically. No "read these 7 files in this order" ritual.
Configurable
Three project sizes (pet, standard, and SaaS) plus host-language profiles: configuration as data, not prose advice.
Agentic
Cascade, pre-flight, and handoff assembly are real tool calls, not instructions the agent has to remember to run.
Drop-in
One command clauderizes any repository, in any language. Idempotent; never clobbers your files.
Host-portable
One install targets Claude Code, Cursor, Copilot, Codex, Gemini, Windsurf, Cline, Amp, Continue, Zed, or kimi.
Maturity
Stable, with receipts.
Every quality gate is publicly verifiable: the test suite passes on Ubuntu, macOS, and Windows across Python 3.11–3.13; the Windows wrapper is executed, not simulated; cold starts are proven on both host shapes; and the quickstart runs against the published package in a clean CI environment on every push.
38MCP tools + 3 resources
3×3OS × Python, all passing
0runtime dependencies
Apache 2.0licensed, explicit patent grant
Markdown is the source of truth. The graph index is a disposable cache rebuilt on demand. If they ever disagree, Markdown wins.
Quickstart
Three steps. Then just drive.
One command clauderizes any repository, in any language. You talk in plain English; your agent makes the tool calls and keeps your plan, graph, and memory in sync between sessions.
- 1Clauderize any repo
- 2Point it at a goal
- 3Then, every session
Prefer a real install? pipx install "clauderizer[mcp]". The core engine has zero runtime dependencies.
Give your agent a memory that survives.
Project memory belongs in the repository, where it ships with the code, shows up in diffs, and is reviewed alongside changes. As tool calls the agent can't forget to make, not advice it has to remember.