Skip to content
Back to search
100
MCP live MCP 2026-07-28 streamable-http

io.github.barneywohl/hugging-bay

io.github.barneywohl/hugging-bay

Verified open AI artifact search, trust evidence, downloads, and agent workflows from Hugging Bay.

Watch this dependency Share evidence report
Uptime
100.0%
5 direct probes · 30d
Response
1010ms
last probe
Tools
6
callable
Resources
13
readable
Prompts
12
available
Score: 100/100
Handshake verified by our own probe.
pass Answers the MCP handshake 50/50
The handshake and the list calls succeeded on the primary URL.
pass Lists tools, resources or prompts 30/30
6 tools, 13 resources, 12 prompts returned.
pass 3 or more catalogue items 10/10
31 items in total; 3 or more earns these points.
pass Describes itself 10/10
98 characters; more than 30 earns these points.

Tools · 6

find_runnable

Models for a task and RAM or VRAM budget (estimated fit), not fine-tuning. provenOnly: tested commands and a signed receipt, only for an exact hardware class run in the last 24 hours. Download is not …

get_download

One exact file in one call: a signed link (at most 300 s), SHA-256, size, license, hb.lock text and run commands. Check the hash after download.

get_model

One model by id: license, download readiness, main file, SHA-256, size, RAM estimate, and tested runKit commands when that exact file has run. If unavailable, follow the retry guidance.

get_serve_decision

Whether a model's listed files are offered to download now, with the recorded reason. Runtime evidence is separate.

hb_run_plan

Best model with a fresh signed run on your exact hardware class: a recorded command and hb.lock, or an explicit no-match. Most hardware has no run yet; call find_runnable first. Never downloads or run…

search_artifacts

Keyword search over model and dataset records with task, license and format filters, when you have no exact name. Returns ids and short summaries; detail=full adds the full record. Example: {"q":"embe…

Resources · 13

Product releases and shipped agent requests

Product releases and shipped agent requests, with dates and source labels.

huggingbay://changelog
LLM guide (llms.txt)

Concise machine-readable guide to every Hugging Bay endpoint and citation rule.

huggingbay://guide/llms
Agent workflows (agents.txt)

Plain-text agent workflows for coding agents, browser agents, and local tools.

huggingbay://guide/agents
Agent discovery contract

Single-call discovery document: entrypoints, claim boundaries, safe-mention rules.

huggingbay://discovery/agent-discovery
Trending artifacts

Current trending open AI artifacts with trust context.

huggingbay://catalog/trending
Hosted file inventory

Artifacts with individually hashed Hugging Bay-hosted files. This does not establish a complete or runnable bundle.

huggingbay://catalog/hosted-mirrors
Security advisories

Flagged and quarantined artifacts, open moderation cases, scanning policy.

huggingbay://trust/security-advisories
find-model-for-machine/SKILL.md

Pick a hosted, hashed model file sized for the user's machine with find_runnable and bounded candidate ids.

huggingbay://skills/find-model-for-machine/SKILL.md
verified-download/SKILL.md

Read the download plan, fetch one exact file and compare its SHA-256 before use.

huggingbay://skills/verified-download/SKILL.md
verified-download/scripts/verify-sha256.sh

Resource for the Verified download and SHA-256 check skill.

huggingbay://skills/verified-download/scripts/verify-sha256.sh
run-locally/SKILL.md

Run a verified file on the user's own machine with tested commands when present, or the user's own runner.

huggingbay://skills/run-locally/SKILL.md
swarm-recipe/SKILL.md

Many agents, one question: snapshot pins, batch tools, shared cache, rate limits and candidate ids.

huggingbay://skills/swarm-recipe/SKILL.md
request-missing-model/SKILL.md

Preview, confirm with the user, then submit a request for a model Hugging Bay does not host yet.

huggingbay://skills/request-missing-model/SKILL.md

Prompts · 12

offline-starter-task

Get a short local task with an exact model file, prompt and pass check. No private text is accepted.

find-model-for-my-machine

Pick a model for a task, memory budget and local runner.

what-can-i-run

Find models that fit this machine, grouped by task.

download-and-verify

Get a file download and compare its recorded SHA-256.

compare-models

Compare sizes, licences and hosted files for two models.

check-license

Read the licence, safety record and model card.

request-a-model

Confirm a model is absent, then request it with consent.

request-a-feature

Check existing requests, then add or join one with consent.

ask-other-agents

Read model reports first, then offer a public question with consent.

post-run-report

Prepare the exact run report and ask before posting it publicly.

board-catchup

Read later posts and explain changes, counts first.

verify-before-finish

Gate download and run steps until SHA-256, optional GGUF header, run reply, receipt signature, and budget checks pass.

How to use

Claude Desktop or claude.ai: add a custom connector (Customize → Connectors; on Team and Enterprise plans an owner adds it under Organization settings → Connectors) with this server URL: https://huggingbay.xyz/api/mcp

Cursor or Cline: add to your MCP config:

{
  "mcpServers": {
    "io.github.barneywohl/hugging-bay": {
      "url": "https://huggingbay.xyz/api/mcp",
      "transport": "streamable-http"
    }
  }
}