devin-usage

Account plan, quota and credit usage for the Devin (Cognition) provider in pi — quota bars in a transcript card plus a compact footer status.

/devin-usage     # full report card (plan, daily/weekly quota, credit buckets)
/devin-models    # per-model credit multipliers (burn rate vs 1×)

A summary also lands in the footer status on session start (devin: Pro · day 82% · wk 91%).

What you get

• Plan: Devin Pro / Teams / Free, org, billing window
• Daily & weekly quota: % remaining with reset countdowns
• Credit buckets: prompt / flow / flex credits used and left
• Model burn rates: every registered Devin model's credit_multiplier (×2 cheap, ×230 premium) sorted cheapest-first, with cost tier and pricing type, flagging your active model
• Overage: any overage balance surfaced when present

How it works

Devin ships no REST usage endpoint; everything comes from one unary Connect RPC the native CLI issues at startup, ported here from oh-my-pi:

POST https://server.codeium.com/exa.seat_management_pb.SeatManagementService/GetUserStatus
Content-Type: application/proto
Connect-Protocol-Version: 1
Body: raw (unframed) GetUserStatusRequest protobuf

The backend gates the response on the Metadata identity tuple — the request must announce itself as the released Devin CLI ("chisel"), not the Windsurf identity pi-devin-auth uses for chat:

ide_name "devin-cli" · ide_type "chisel" · ide_version "3000.6.2"
extension_name "chisel" · extension_version "3000.6.2"

Caveat: this is an undocumented internal endpoint guarded by a client identity check. It works today; Cognition could change or gate it at any time, at which point the command will fail loudly until updated.

Requires the Devin provider to be signed in (/login devin) — the API key comes from pi's own model registry, nothing is stored by this extension.

Install

pi install git:github.com/betim-hodza/devin-usage

or try it once without installing:

pi -e git:github.com/betim-hodza/devin-usage

License

MIT