Access · how to use bucket.foundation

One canon. Many surfaces.

Bucket exists in three layers — a website (this page), an open protocol (the x402 rail), and four GitHub repos (the source). End users read the canon. Researchers cite it. AI agents query it. Everyone accesses the same data.

1 · Website (humans)

/canon7-branch grid + interactive globe — entry point /canon/searchType a question, get ranked claim cards /canon/claimsAll 599 curated claim cards /canon/bridges17 multi-branch primitives (cross-domain isomorphisms)/canon/graphKnowledge graph: 1,133 nodes, PageRank rankings /canon/timelineGlobe through time: 50 events 570 BCE → 2020 CE /canon/[branch]Per-branch page (e.g. /canon/physics)/protocolx402 paid-once-cite-forever spec /cite-forever/v0.1Citation license /manifestoWhy bucket exists /governanceNonprofit governance + COI disclosure

2 · API (AI agents · scripts · downstream tools)

/api/canon/searchtry ↗

Search canon claims by query

/api/canon/search?q=consciousness&top_k=5

/api/researchtry ↗

Paid x402 research over feed402 (PubMed/Semantic Scholar)

/api/research?q=mitochondrial+function&tier=insight

/llms.txttry ↗

Playbook for AI agents (the LLM-readable spec)

/llms.txt

/.well-known/feed402.jsontry ↗

feed402 discovery manifest

/.well-known/feed402.json

/.well-known/mcp.jsontry ↗

MCP server manifest

/.well-known/mcp.json

3 · MCP servers (Claude Desktop · Claude Code · any MCP client)

One MCP server, seven tools. Connects local canon files AND the bucket.foundation research API in one server. (Previously split across two repos — consolidated.)

bucket-mcp

Tools: canon_search · canon_get_claim · canon_list_branches · canon_list_bridges · canon_get_bridge · bucket_research · bucket_cite

# 1. clone the repo (one time)
git clone https://github.com/bucket-foundation/bucket-foundation

# 2. register the MCP server
claude mcp add --scope user --transport stdio bucket \
  -- python3 $(pwd)/bucket-foundation/mcp-server/bucket-mcp.py

4 · GitHub (source · contribute · fork)

The GitHub org is github.com/bucket-foundation. Three repos in that org, plus one upstream gateway in the AGFarms org.

bucket-foundation/bucket-foundationgithub ↗

Main site + canon data + tools

The Next.js app you're using. Canon claim cards (bucket-canon/), detected bridges (bucket-canon/_bridges/), embeddings + trained ML artifacts (_intake/), MCP server (mcp-server/), pipeline tools.

AGFarms/x402-research-gatewaygithub ↗

Upstream feed402 gateway (the rail)

The x402-protected research API behind /api/research. PubMed, Semantic Scholar, OpenAlex, ClinicalTrials, PubChem, Kruse corpus. feed402/0.2 compliant.

5 · Research drive (PDFs · raw data · mirror)

BucketDrive

Public Google Drive folder mirroring bucket-research repo. PDFs (which don't belong in git), large datasets, Figma exports, scanned manuscripts. Same canon-branch tree as the repo.

Quickstart by role

Reader

Start at /canon. The 7-branch globe is the map. Click any branch, browse claims. Try search for a specific question.

Researcher

Clone bucket-foundation/bucket-foundation. The canon claims live in bucket-canon/. The data pipeline is in _intake/. See REPRODUCE.md.

AI agent

Read /llms.txt. Hit /api/canon/search or /api/research. Or install the MCP server. No wallet, no key, no auth — anonymous rate-limit only.