Connect Qianye to your Agent

Give your Agent the instruction.

For Agents that can read websites or make HTTP requests. Paste the instruction, then tell your Agent which industries you follow.

01
Set your scopeCategories, markets and your position in the value chain.
02
Read the briefingsYour Agent reads articles, sources and revisions, then checks business implications.
03
Make it a routineSet a schedule in your Agent host to keep up with new publications.

Public reads. No API key.

View the full instruction

Add Qianye to my industry research.
Discovery: https://www.cfo.cafe/agent/manifest.json
Language: en

Confirm my categories, markets and business position, then look up categories in the catalog. Read changes with kind=daily and mode=live; follow each ref for the fixed article revision and its sources. Report when published content is unavailable.
Prepare relevant changes, cost and cash implications, questions to verify and original sources. Process contracts, inventory and budgets locally.
Persist content and events before advancing next_cursor, draining every has_more page. Reuse the same scoped cursor to handle new, corrected and withdrawn content. For daily scheduling, use the host's supported task feature with my authorization.

Protocol and versionr1 · 2026-10-29

Connect Qianye to your Agent

Start a baseline

Read /api/delivery/v1/manifest and catalog. Choose category, language and live/demo scope. Drain changes and read each fixed release_key revision. Public reading does not require private enterprise uploads.

Continue daily

Persist events, text and sources before advancing next_cursor. Continue while has_more is true. Reuse the same scoped cursor the next day; restart when scope changes. no_change is not a new research result.

Handle corrections and withdrawals

Retain stable_id, locale and revision in references. Stop treating a withdrawn edition as current evidence while retaining historical citations. The enterprise or platform configures its own seven-day schedule.

Measurement boundary

Fetching, downloading, reading and polling do not establish adoption. Real research usage remains subject to the independent receipt protocol, deduplicated by workflow_id × research_run_id. This CMS does not convert traffic to traction.

View the published protocol →