The Problem

Three weeks in, nobody knows how it works.

Building got fast. Keeping track did not. What’s done, why it was built that way, and what the next change will break are buried in chats nobody opens again.

It works, and nobody knows how.

The AI built it, and how it works was never written down. When it breaks, every fix is a guess, until starting over feels easier.

Every new chat knows only half the story.

It reads the handover and says it’s ready. Then it builds on the half it knows, and you find the missing half when something breaks.

You told it. It did the other thing.

The rule was right there, and it went the other way. So you write it again, louder, in capitals, and add three more rules to be safe. Next session, it ignores those too.

Every change feels like a gamble.

You say “go ahead” and hope it understood. A week later you find what it really built, and what it broke that you never knew was connected.

The fix is not a better assistant, and it is not another folder. Grasppy is a system — it has steps, and they run while the work happens. Not afterwards, when nobody has the time.

AI memory, automated

One shared memory across every AI tool, updated every minute.

Start in Claude Code, continue in Codex, and nothing needs explaining. Every session feeds the same memory, so any chat already knows what was decided and why. No handoff prompts to write: Grasppy keeps writing them for you while you work.

One memory

Context compounds

Every turn, in every tool, adds to it.

01 · Capture

Every session kept

Claude Code, Codex and Cursor sessions are captured as you work.

02 · Notes

What was decided

Each turn's decisions and fixes become notes, linked to the turn. Every 12 turns condense into a chapter.

03 · Recall

Shared with every chat

Switch tools mid-task: the next chat already knows the last decision. Nothing to paste, nothing to explain.

04 · Update

Stays current

When a decision changes, the old note is marked as replaced automatically. Override it in one click, if you ever want to.

05 · Continuous

Never stops running

Every turn, mid-conversation or across tools, feeds the shared memory, with nothing to start or stop.

The longer you build, the less you explain, and your next chat starts where the last one ended.

My Code keeps a live map of your project.

Modules, files and what depends on what, refreshed automatically as the code changes, so every chat knows what a change will break before it edits. Your source code never leaves your machine.

Context, recalled

Everything your local AI needs, one folder away.

Every turn is summarised as it happens, so the problems solved and decisions made sit beside your plans, docs, memory, research and code map, all in one folder in your project. Recall's notes update about every minute, the folder refreshes every five minutes when something changes, and your local AI finds what it needs on its own, with no API calls and no pasting.

Not a dump of files, a map. CONTEXT.md tells your local AI where everything is, indexes point it to the turns that matter, and the graph shows how it all connects, so it reads the few right files instead of everything.

Always current Recall's notes update about every minute; the folder and code map refresh every five minutes when anything changed. When nothing changed, nothing runs.

Read by any local AI Claude Code, Codex and others start at CONTEXT.md and follow the map.

See how it connects Conversations, decisions, docs and code as one interactive graph.

Safe by design Grasppy writes the files; nothing you edit on disk can overwrite your workspace.

Explore Context Vault Built for local AI, not for storage.

grasppy-exports/ synced 3 min ago

context-vault/CONTEXT.md root map

context-vault/analysis/ decisions & fixes

mydocs/ documentation

myplans/ plans

mymemory/ memory

myresearch/ research

mycode/ live code map

105_Claude-Mem · live notes updated every turn

Chapter summary

Notes, turn by turn

Investigated

The page downloaded the whole project when boxes were ticked
Large projects went over the 6 MB reply limit

Learned

Load only what is on screen; exports fetch their own sections

Completed

Pagination, batch loading and compact replies, shipped in 9.131

Next steps

Window the tree once a project passes 5,000 nodes

Codex · new chat

YouWhy does the docs page no longer download the whole project?

Local AIopens CONTEXT.md → myplans/ → Plan 38904

Large projects froze the page, so exports now load only the selected sections.Decided Oct 8 · source: the exact turn

Analysis, in depth

Not just captured. Analysed, turn by turn.

Every turn gets its own summary as it lands, and its decisions, problems, fixes and the files they touched are pulled out and linked to the exact moment. Press Analyze, and the whole conversation becomes summaries, diagrams and a map of how everything connects.

As you work Every turn, as it lands A summary of every turn, its notes, and a chapter every 12 turns. Live Automatic

On demand The whole conversation Summaries, diagrams and relationships, built together so they agree. Analyze One click

Seven-stage analysis

Extract, enrich, connect

A conversation stops being a scroll and becomes a structured record: what was decided, what broke, what was built and how it all connects.

Decisions & fixes

Typed notes, turn by turn, each linked to its moment.

Artifacts

Every file the AI wrote, versioned across the session.

Chapters

Where each stretch of 12 turns actually ended up.

Summaries & briefs

Executive summary, technical brief and a handover prompt.

Diagrams

Decision trees and process flows, drawn from the conversation.

Relationships

Topics and entities, connected on the semantic map.

Everything links back to the exact turn it came from, so any claim can be checked in one click.

Inside Grasppy

The real thing.

Real screens from a real workspace. Every number, chart and chat below is Grasppy in use.

Your whole project, at a glance

Everything captured across your AI tools in one place: analyses, documents, artifacts, plans, prompts and skills, with your recent activity beside them.

Grasppy dashboard showing 9 documents, 114 analyses, 12 artifacts, 13 projects, 16 prompts, and 23 skills — with a 5-week activity calendar, weekly analyses bar chart, and a platform-breakdown donut chart

Every conversation, restructured

Press Analyze and a conversation becomes summaries, decision trees and process flows, every node linked back to the turn it came from.

Grasppy Decision Tree view rendering a color-coded flowchart of decisions and reasoning extracted from an AI conversation, with the original chat thread visible side-by-side for source verification

Every project, in one tree

Plans, documentation, rules, skills, prompts and memory as versioned projects in one workspace, synced to a folder your local AI reads.

The Grasppy Context System — an always-on context layer feeding the compounding work loop

The whole project as one map

The Semantic Map connects conversations, decisions, documents and code, so you can start at an idea and land on the exact turn where it was decided.

Grasppy Knowledge Graph view in dark theme showing Semantic Intent Markers — a dense web of color-coded nodes and connections mapping the entities, decisions, and concepts across an AI conversation

Who It's For

Built for vibe coders, and every project that outgrew its first prompt.

Whether you write the code yourself or the AI writes most of it, the trouble arrives the same way: the project grows past what anyone can hold in their head.

★

🚀

Vibe Coders

You build by describing, not by typing every line, and that's fine. Grasppy keeps the record you don't: what was built, why, and whether it matches the plan, so the project stays yours even when the AI wrote most of it.

👨‍💻

Developers

Engineers shipping production code with AI — Claude Code, Codex, Cursor, Windsurf, Copilot. Keep every architecture decision tied to the code it produced, and know when the two stop matching.

👥

Dev Teams

Shared memory for the whole team: every seat works from the same decisions, plans and docs, with access set per workspace. Seats come with Pro and Max.

The tools that write your code do not remember writing it. Grasppy does — and checks what got built against what you planned.

Frequently asked questions

Everything you need to know to get started.

What is Grasppy?

Grasppy is the workspace around your AI coding tools. It captures every Claude Code, Codex and Cursor session as you work, analyses it turn by turn into decisions, fixes and notes, and keeps your plans, docs, rules and memory current, so any chat, in any tool, starts already knowing your project.

Your local AI reads all of it from one folder, plans are checked against your actual codebase, and nothing changes your docs, rules or memory until you approve it.

Is Grasppy available now?

Yes — Grasppy is live. Start your free trial today: 14 days on the Light plan. Cancel anytime before it ends and you won’t be charged.

Which AI tools does it support?

17 integrations. Browser tools via our Chrome extension: Claude, ChatGPT, Gemini, Grok, DeepSeek, Perplexity, Mistral, GitHub Copilot, TypingMind, Replit, Bolt, Lovable, and Voila.

Coding tools, captured automatically as you work: Claude Code, Codex Desktop and Cursor. Windsurf through Markdown import.

If your tool exports conversations as Markdown, you can import them today.

Do I need my own API key?

No. AI is included in every plan. Grasppy runs the models and pays for every AI call, and your plan's monthly credits cover the work. There is no provider account to create and no key to paste — sign up and analysis simply works.

What does it cost?

Three plans, AI included: Light $29/mo (1 workspace, 3,000 AI credits a month), Pro $49/mo (3 workspaces, 2 seats, 10,000 credits) and Max $99/mo (unlimited workspaces, 3 seats, 20,000 credits). Need more? Top up at $5 per 1,000 credits; top-ups never expire. See the pricing page for the full comparison.

Capturing, reading, searching and organizing your conversations never uses credits. Only AI work, such as analysis and Ask, does.

What is Grasppy Studio, and is it really included?

Grasppy Studio is a content-creation suite built on top of Grasppy. It includes Voice Calibration across 5 dimensions, AI writing through your own local AI agent (zero API cost), a 6-stage SEO surgical-edit pipeline, keyword research via DataForSEO, and a built-in blog with a custom URL.

It's included with your Pro subscription. See the Grasppy Studio landing page for the full feature list.

How is Grasppy different from Notion, Obsidian, or ChatGPT Projects?

Notion / Obsidian are blank pages — you have to write and organize everything yourself. Grasppy does that work for you automatically from your AI conversations.

ChatGPT Projects / Claude Projects are locked to one platform and don't extract decisions, link artifacts, or build a searchable knowledge graph across tools. Grasppy works across all your AI tools and turns chat into structured, reusable knowledge.

Is my data secure, and does Grasppy train AI models on it?

Secure, yes. Trained on, never. Your conversations are yours. Grasppy does not train models on them, sell data, or share content with third parties.

Data is encrypted in transit and at rest. AI processing runs through Grasppy's own accounts with its AI providers, only to do the work you asked for.

Can I export my data if I leave?

Yes. Every conversation, document, and artifact can be exported as clean markdown at any time — no lock-in. Your knowledge belongs to you, in a portable format you can take anywhere.

Is there a free trial?

Yes. Every new account starts with a 14-day free trial of the Light plan, with about 1,000 AI credits included. Cancel anytime before it ends and you won’t be charged; otherwise it continues on Light, and you can switch to Pro or Max whenever you like.

I'm not a developer. Can I still use Grasppy?

Yes, and that's exactly who Grasppy is built for. If you build with AI by describing what you want, Grasppy keeps the record of what was built and why, and checks it against the plan, so you stay in charge of a project you didn't type line by line.

It works just as well for anyone running long AI conversations: researchers, writers, consultants and product managers.