You are not losing context.
You are losing continuity of thought.
OCH is a small, human-readable Snapshot format for carrying working context from one AI system to another.
If you switch between ChatGPT, Claude, Gemini, Cursor, Claude Code, or other agents, OCH helps you continue the same work without rebuilding the whole story.
No install. No account. No agent framework. Just a reviewed handoff artifact you can copy between AI systems.
Star OCH if you want:
- a copy/paste protocol for cross-AI handoff
- prompts that force one clear
NEXT ACTION - a human-reviewed context artifact, not hidden platform memory
- examples and trial templates for testing whether handoff actually worked
Do not star it expecting a CLI, agent, memory system, or automation framework. OCH is intentionally a protocol.
- Open USE_OCH.md.
- Copy the source prompt into your current AI window.
- Review the generated Snapshot.
- Paste the receiver prompt and reviewed Snapshot into a new AI window.
For the guided path, start with START_HERE.md. For the full trial, use docs/15-minute-trial.md.
| Need | Start here |
|---|---|
| Fastest copy/paste path | USE_OCH.md |
| Snapshot format | docs/snapshot-format.md |
| Generator prompt | prompts/snapshot_generator_prompt.md |
| Receiver prompt | prompts/receiver_prompt.md |
| Real examples | examples/ |
| Trial result template | tests/user-trial-result-template.md |
Without OCH, a new AI window often starts with:
Can you share the project background and what has already been decided?
With OCH, the receiving AI can start from the handoff:
I’ll continue from NEXT ACTION while preserving the stated DECISIONS and CONSTRAINTS.
The handoff artifact is a reviewed Snapshot, not the full prior conversation.
### WHAT WE ARE DOING
...
### DECISIONS
- Do not reopen the settled direction.
### CONSTRAINTS
- Keep the output human-reviewed.
### NEXT ACTION
Continue from the next concrete step.We no longer work with a single AI system.
We constantly move between:
ChatGPT · Claude · Gemini · Cursor · Claude Code · Agents
But something breaks every time we switch:
Our thinking does not carry over.
Only fragments do.
Every switch forces you to:
- re-explain what you're doing
- rebuild your decisions
- re-establish what is true
- reconstruct the same narrative
Meanwhile:
AI quietly rewrites your story in its own way.
Sometimes correct.
Sometimes confidently wrong.
You are no longer continuing your work.
You are restarting your thinking.
There are two types of AI users:
They treat every new AI window as a blank slate.
They preserve context across systems and continue from where they stopped.
Not to store memory.
Not to automate workflows.
Not to build agents.
But to solve one thing:
Make thinking portable across AI systems.
OCH generates a Context Snapshot:
A small, human-readable artifact containing:
- what you are working on
- what you have decided
- what you explicitly rejected
- what you should do next
You can carry it across:
- AI chat windows
- different models
- coding agents (Cursor / Claude Code)
- personal notes
And continue with less re-explaining.
For the current V1 first-user path:
- Start with the practical guide
- Review the Snapshot v1 format
- Copy the Snapshot generator prompt
- Copy the receiver prompt
- Record the trial result
- End a session while thinking is still coherent
- Generate a Context Snapshot
- Review and approve it
- Paste into another AI system
- Continue from NEXT ACTION
AI drafts.
Humans decide.
Context belongs to the user.
AI systems are becoming more powerful.
But human thinking is becoming more fragmented.
We optimize models.
But we lose continuity.
OCH is a small attempt to fix one thing:
Keep thinking alive across system boundaries.
OCH is intentionally narrow.
It is NOT:
- a memory system
- an AI agent
- an automation tool
- a knowledge base
How do you continue thinking without restarting yourself every time you switch systems?
Context does not belong to platforms.
It belongs to the person thinking.