sessionbin

A transcript pastebin for agentic sessions.

Upload a session file, get a shareable URL, and anyone with the link can view the server-rendered transcript. Anonymous upload, delete-by-token. No accounts, no dashboards.

Currently supported: Claude Code, Codex, OpenCode, and Pi. Future transcript support for: Gemini CLI.

How it works

1. Upload a raw session file (via web UI or API).
2. The server redacts secrets, parses the session, renders it to HTML, and stores it.
3. You get a share URL and a delete token.
4. Anyone with the URL sees the rendered transcript. Only you (with the token) can delete it.

The companion CLI lives in cli/ in this repo and is published to PyPI as sessionbin.

Local development

Everything needed to run and exercise the project lives in a container: the backend, the CLI, gitleaks, and all supported agent harnesses so you can produce real sessions.

podman build -t sessionbin-dev -f dev/Containerfile dev/
podman run --rm -it -p 8000:8000 -v "$PWD":/repo:Z \
    -v sessionbin-uv-cache:/root/.cache/uv \
    --name sessionbin-dev sessionbin-dev

The site is then at http://127.0.0.1:8000. The container syncs dependencies, installs the CLI from cli/, applies migrations, and runs the dev server against your mounted checkout, so edits reload live. The database and stored pastes go to /state in the container.

To produce a real session and upload it, exec into the running container:

podman exec -it sessionbin-dev bash
cd /work && mkdir -p demo && cd demo          # any scratch directory
export ANTHROPIC_API_KEY=...                  # or pass --env-file to podman run
export OPENAI_API_KEY=...                     # for Codex and Pi

claude -p "..." --model claude-haiku-4-5      # Claude Code
codex exec --skip-git-repo-check "..."        # Codex
opencode run -m anthropic/claude-haiku-4-5 "..."   # OpenCode
pi --provider openai --model gpt-5-mini -p "..."  # Pi

sessionbin upload --latest -y --server http://127.0.0.1:8000

Run the checks with tox run (inside the container or on the host).

Adding support for a new harness? Install it in dev/Containerfile as well, or the dev container can no longer produce a real session to test the new adapter against.

Running the backend outside a container needs Python 3.14 and gitleaks on PATH — the settings module raises at import without it.