The control plane
for your AI coding agents.
Every agent, every token, every dollar. On your machine.
pip install superbased-observercode --install-extension superbased.superbased-observer/plugin marketplace add superbasedapp/pluginsSee, control, and right-size every agent from one binary.
Provider-reported tokens and cost across all 40 tools. Cache read vs write. A next-message cost predictor before you send.
Launch 20+ CLI agents as dashboard terminals. Jump into a live session, take the keyboard, hand it back. Reach it all from your own tailnet.
Route models by task. Gate egress before a command runs. Cut the bill you didn't mean to run.
Thirty days of agents. No idea what it cost.
dashboard · localhost:8081
Every tool keeps its own session files. Nobody adds them up.
Start an agent. It opens as a terminal.
0 / 4 SESSIONSFour agents, four repos, one screen. Attach to any of them mid-run, over your own tailnet.
Watch your coding agents and the app you ship to users.
Local capture across 40 tools, per-developer or rolled up for a team. Cost, sessions, routing, guardrails. Nothing leaves the box unless you turn on org sharing.
General observability for a hosted LLM app: OTLP traces, evals, and input admission that can hold a request before it ever reaches the model.
Step five is about to cost you forty dollars.
One keystroke, 45x cheaper, because you could see it coming.
See last month's spend in about nine seconds.
The same prebuilt binary ships through npm, PyPI, and the VS Code Marketplace. Claude Code users add the plugin from the marketplace. No account, no config.
npx @superbased/observernpm install -g @superbased/observerobserver startpip install superbased-observerobserver startcode --install-extension superbased.superbased-observer/plugin marketplace add superbasedapp/plugins/plugin install superbased@superbased