Pantheon
Change the agent.
Keep the conscience.
Continue your work in any agent with the conversation, files, and decisions intact — stored in a private GitHub repository you own.
Works with
- Cursor
- Claude
- Codex
- Windsurf
- and desktop clients with local MCP
Use Pantheon when you need to
Change models mid-project
Move from Cursor to Claude, Codex, or another desktop agent without explaining the work again.
Start a clean conversation
Leave a tired context window without losing decisions, failed approaches, or files.
Hand work to someone else
Share one read-only conscience instead of writing a summary that drops the evidence.
Stop repeating corrections
Turn recurring project lessons into reviewed skills the next agent can fetch when needed.
Context continuity
The model can change. The project state should not.
Pantheon gives desktop agents a continuity layer outside any one context window: the conversation, decisions, files, and reviewed lessons persist, while each new agent loads only what the current task needs.
Long context stores more tokens. Context continuity carries forward the work.
Survives a fresh chat
Leave a tired context window without losing the decisions and evidence that made the work coherent.
Survives a model switch
Resume the same conscience in another desktop agent through a common MCP tool surface.
Keeps the summary reopenable
The next agent starts with a bounded packet, then searches the underlying turns and files when it needs proof.
Agents forget. That is the whole problem.
Every chat starts at zero
An agent remembers nothing of the conversation you had with it yesterday, or the one running in the next window. Whatever you explained, you explain again.
The window ends before the work does
Long sessions degrade before they finish. You open a fresh one to get the quality back, and the reasoning that got you this far does not come with it.
A summary is not the evidence
It is a description of a conversation, written afterwards, by the participant with the worst memory of it. The command that failed, the output that explained why, the file that was finally right — none of that survives the retelling.
Code is the cheap half now
There has never been more supply of code. What stays scarce is why it looks like that: the three approaches you rejected, what broke last time, what the customer actually said.
How it works
Two commands, typed to your agent rather than into a shell. Everything between them is a git commit you own.
- 01
Hand off where you are
Pantheon reads the session file your agent already keeps on disk — the full conversation and the tool activity that the client records. It does not ask the model to retype the conversation, and it refuses a recap pretending to be one.
/pantheon-handoff my-project - 02
Your context stays in your repo
Every saved conversation, file, decision, and memory is written to one private GitHub repository you own. Pantheon stores no conversation content in its database and receives no access to your other private repositories.
- 03
Resume in any other agent
The next agent gets a short packet to orient itself, plus a map of everything the space holds — so it can open an earlier turn, search a tool call, or pull a stored file when it needs the detail rather than the gist.
/pantheon-resume my-project
And this is all it is
No proprietary format and no export button to find later. Your private repository is the source of truth; each space is a folder of readable files and every handoff is a commit you control.
It notices what you keep
having to say.
You told the last three agents not to add shadows. Each time the correction landed, worked, and died with the chat — because a conversation ends and nothing carries the lesson forward.
When a handoff adds turns, the agent reviews repeated corrections, recurring procedures, and failures that hit twice. It reads the evidence and batch-saves justified lessons as inert drafts for you to approve later. Run the command below to review again at any time.
/pantheon-skills my-projectWaiting on you
no-shadowsAwaiting review3 sessionsUse when adding elevation to any surface in this project.
This project does not use drop shadows.
Flat surfaces are separated with a one-pixel border, never elevation. This came up on the board tab and again on the mockup, and was reverted both times.
Drawn from cursor-a91264d3…#482, cursor-a91264d3…#1435. Every id was checked against the vault before this was saved.
See a correction become a reusable skill.
Pantheon finds the repeated signal and its evidence; the agent drafts the lesson, and you decide whether every future agent should receive it.
- You approve every one
- The review writes drafts without interrupting you. None reaches an agent until you promote it on the space page; reject the rest there.
- Every claim is sourced
- A skill names the turns it came from, and the vault refuses to save a citation that does not resolve. You can open the moment and check it.
- Twelve, at most
- Each live skill's name and trigger ride every resume, so even lazy loading has a context cost. The cap is the feature — a list nobody reads is worse than no list.
- Then every agent has it
- Promoted skill triggers arrive with the resume in whichever model you open next. The body is fetched only when the work matches.
Hand someone the thread,
not a summary of it.
Onboarding someone, handing over a project, getting a second opinion on a bug you have been inside for two days — the honest answer is always “read this conversation”, and there was never a way to do that.
Give one space to another Pantheon user and they get the whole conscience: every archived conversation, the decisions, the files. Their agents resume it by name, in whichever model they use. Not a link to a chat — a conscience their tools can read.
/pantheon-resume obirikan/preordersTheir consciences
safari-login-loop
4 sessions · 2 files
preorders
Read-only. Resume it, search it, open any turn — but a handoff aimed here is refused before anything is written.
- One space, never the repo
- You pick the conscience to hand over. Everything else in your vault stays invisible — they are given a folder, not an account.
- Read-only by construction
- A handoff aimed at someone else's space is refused where the path resolves, not hidden behind a disabled button. Your conscience stays yours.
- It works in their agents
- Their CLI pulls it alongside their own spaces, so it resumes by name in Cursor, Claude Code or Codex — not only in a browser tab.
- Revoking reaches their disk
- Access ends at once and their next sync deletes the local copy. What they hold is a cache of a permission, not a gift you cannot take back.
Start with one conversation.
Create one private vault in your GitHub account and connect Pantheon only to that repository. Your context remains readable, portable, and under your control.