01 · Onboarding
Docs → spec in 3 minutes
Drop a PDF, form screenshot, or FAQ doc. Five Claude stages extract every field, rule, and routing path with a verbatim source quote. No hallucinated rules.
Solutions
Kopi is a Lark expense approval bot that connects natively via webhook and sends AI-checked result cards directly to approvers. Whether your team uses Lark or Feishu (飞书), Kopi's Feishu expense approval integration requires no new tools and no behavior change for employees.
Enter your Lark App ID and App Secret in Kopi's onboarding wizard. Kopi verifies the connection and retrieves your workspace's approval flow definitions.
Choose which Lark approval flows to monitor. Kopi registers a webhook subscription — employees continue submitting expenses exactly as before.
On each submission, Kopi downloads attachments, runs 5-dimension AI analysis (category, amount, date, invoice, rules), and delivers an interactive check card to the approver in Lark.
The approver taps green, yellow, or red in the Lark card. Confirmed categories are saved to the rule library — same item next time is auto-approved.
No credit card. No time limit. Shape the product with direct founder access.
Built on Mirror
Lark is one of four delivery surfaces Mirror supports — alongside web forms, email, and Slack. Already-connected tenants get card-style approvals; new tenants can launch on web + email and add Lark later.
01 · Onboarding
Drop a PDF, form screenshot, or FAQ doc. Five Claude stages extract every field, rule, and routing path with a verbatim source quote. No hallucinated rules.
02 · Runtime
runTemplate(spec, request) → Decision is pure code, no LLM at decision time. The spec is data, the runtime is deterministic. Three-color verdict with full routing trace.
03 · Channels
A submission can come from anywhere. Reviewers see the same triage card across all surfaces. Web + email is the default; Slack and Lark are accelerators for already-connected tenants.
Citation contract — non-negotiable
Every field, rule, and routing approver Mirror emits carries a verbatim quote from your source document. If she can’t cite it, she drops the item. No fabricated policies — your AI is anchored to your own words.