The other output of coding with AI
The dead end. The useful tool call. The moment it clicked. Turn your AI coding sessions into something the whole team can learn from.
Your infrastructure. Your sessions. No cloud account required.
Search got slower. Do we need a new database?
EXPLAIN QUERY PLAN
SELECT … WHERE session_id = ?;
SEARCH nodes USING INDEX
session_id_ordinal (session_id=?)
Match the index to the filter and sort.
Keep the query bounded. Keep the stack small.
Claude Code · Codex · pi
Supported structured export formats
One self-hosted application
Go + SQLite + Docker
Readable without JavaScript
Server-rendered, bounded transcripts
A pasted answer loses the process. A giant transcript loses your attention. Seshshare makes it easier to find, follow, and share the part that matters.
Link to a session, a nested tool call, or its result. Readers arrive with surrounding context, not a wall of chat to search through.
Keyword search runs inside your instance. Jump from an error, function, or idea to the entry where it appeared.
Vote sessions up or down. See source-reported token and cost estimates, with missing data called out instead of guessed.
Local first. Not local later.
Run the platform on your own infrastructure. Your sessions don't need to pass through someone else's cloud to become useful to your team.
# On the machine running your harness seshshare-publish prepare \ --input selected-session.jsonl \ --out review.json # Review locally. Then publish those exact bytes. seshshare-publish publish \ --file review.json \ --sha256 REVIEWED_HASH \ --instance https://sessions.internal
No automatic history upload. No implicit destination.
Review and redact locally before publishing.
Make the next session a little smarter
Start a shared library on infrastructure you already trust.