TitpeaRemembers your storage
Accounts, buckets, and open Issues, most severe first.
Storage Agent · Looks after object storage
Your S3, looked after.
Read-only tools. Local keys. Issues it remembers.
Findings outlast the task.
Fix packs
An AWS CLI command, a Terraform resource, or the API document.
Fixes you apply
Generated text you run with your own credentials.
Verify, read-only
A re-check that can resolve the Issue.
Opt-in watch
Re-checks an account on a schedule you set.
Read-only tools
No writes, deletes, or policy changes.
Where your data goes.
- Storage credentials stay in an encrypted vault on this machine and are masked from tool output.
- Evidence and task history stay in the app’s local data directory.
- Requests go to the model and S3-compatible endpoints you configure.
Sep 30, 2026
What’s new in v10.0.0.
- The first request fits an 8k window
- Long turns stay inside the window
- 15 tools instead of 30
FAQ
Does it need a model key?
Not for a local model: Ollama, LM Studio, vLLM, and llama.cpp run without one. Hosted providers such as OpenAI, Anthropic, DeepSeek, and OpenRouter need their key. A local model with no context window set is treated as 16,384 tokens; set a larger value in Settings if your server runs one.
Will a small local model work?
Yes. A local model with a 16k window can finish a first survey, and its first request fits in 8k. With Ollama, the app sends the planned window as num_ctx. Test in Settings makes one real tool call and warns when Ollama’s context is smaller than planned.
What does the watch do?
Nothing until you turn it on for an account; it is off by default. On its schedule it runs a bounded, read-only sweep and opens one task only when something new turns up and a model is configured.
What happens when I upgrade from v4?
Upgrading from v4 imports your model and storage settings, known buckets and Issues, and past results on first start. The v4 database is left untouched.
What will a fix change?
Each fix comes with a preview drawn from evidence the app holds, such as anonymous requests in the bucket’s access logs or existing lifecycle rules. When the evidence cannot tell, it says so.
How is it different from StorageOps?
StorageOps adds storage skills to Pi. Storage Agent is the full desktop app: tasks, read-only tools, local analysis, and the Issues it remembers.
Storage Agent