OnlyFourBot Trusted source network
For people, agents, and reusable AI work

可信关注源网络,不是又一个新闻摘要站。

OnlyFourBot lets the community publish verified source collections, official digests, and machine-readable updates that can be reused by humans, MajorRSS, and agents.

Sources are shared. Private prompts and API keys stay with the user.
AI Infrastructure Radar 96 Trust
Shared sources human view
OpenAI Blog
official low noise 1.8k subs
Hugging Face Blog
healthy models infra
GitHub Changelog
verified developer tools
Official digest

Five verified updates, one disputed claim, zero private content.

Readable summary for lightweight users with source links, confidence labels, and rights boundaries.

Agent endpoint machine view
// MCP resource
onlyfourbot://collections/ai_infra

{
  "cursor": "cur_8x41",
  "confidence": "verified",
  "rights": "public_summary",
  "updates": 5,
  "rate_limit": "read_only"
}
128 personal digests and agent workflows reused this collection this week.

共享的是可信源,不是每个人的私有判断。

OnlyFourBot keeps the public layer clean: verified sources, collections, health scores, rights policies, and official digest. Users and agents build their own analysis on top.

S

SourcePreset

Stable source identity with canonical URL, trust score, health score, noise level, rights boundary, and reuse metrics.

C

Collections

Community-created source bundles that humans can follow and agents can subscribe to by stable ID.

D

Official digest

Low-frequency public summaries for users who just want the signal without running their own agent.

R

Rights policy

Every source declares whether it can be summarized, quoted, redistributed, cached, or used only as a link.

A

Agent-ready by default

Cursor, ETag, since, MCP resources, REST API, signed requests, and scope-limited access are part of the product contract.

一次创建,持续复用。

A good source list, digest, diff, or analysis should not disappear after one token spend. It becomes a reusable asset that can power readers, workflows, and future agents.

01 Create MajorRSS users create private sources, collections, and digest outputs with their own API keys.
02 Sign The client signs public submissions so the platform can verify that sources came from the app.
03 Review Rules, AI risk review, health checks, and community feedback protect the shared network.
04 Publish Trusted sources and digests become discoverable with facts, confidence labels, and rights policies.
05 Compound Humans read the official digest. Agents reuse the source graph to create their own outputs.

一套底层对象,两种消费界面。

Humans need context, trust, and a clean daily view. Agents need stable IDs, schemas, scopes, and predictable update cursors.

For humans

Explore collections, read official digest, inspect trust labels, subscribe with one click, and report noisy sources.

  • Source pages explain health, noise, rights, creator trust, and recent updates.
  • Digest pages show observed, verified, disputed, and low-confidence claims.
  • Creator pages make contribution quality visible without turning the site into a content farm.

For agents

MCP, REST, CLI, and webhook surfaces expose the same network as structured data with rate limits and scopes.

  • onlyfourbot://collections/ai_infra_radar
  • GET /v1/collections/{id}/updates?cursor=cur_8x41
  • Scopes: read:sources, read:digests, subscribe:collections

公共层可信,私有层自由。

Users can privately follow any source. Public sharing is signed, reviewed, scored, and continuously monitored. BYOK keeps personal model cost with the user, while official digest stays limited and cache-first.

BYOK Private summaries use user keys.
Read-only Agents start with safe scopes.
Cached Public facts reuse fingerprints.
Rights Paid reports are links, not redistributed files.