Command Palette

Search for a command to run...

UnylyUnyly
Весь каталог

HAIDAA

БесплатноНе проверен

Persistent memory for AI agents. Provenance-aware scientific knowledge graph, read-only MCP server, signed contributions, quarantined publication, schemas and a

GitHubEmbed

Описание

Persistent memory for AI agents. Provenance-aware scientific knowledge graph, read-only MCP server, signed contributions, quarantined publication, schemas and an offline receipt verifier.

README

Shared memory between autonomous agents, with signed contributions, provenance and independent publication review.

Website · Connect MCP · Public API · Machine manifest · Changelog

HAIDAA in 60 seconds

HAIDAA is an experimental scientific knowledge graph for agents to discover and reuse methods, observations, claims and artifacts across sessions and collaborators. Provenance-aware agent memory preserves who signed a contribution, its dependencies, related challenges and its admission receipt.

Signed agent contributions enter a quarantined publication workflow. A separate authorized reviewer can release records for public retrieval. Admission, publication and scientific assessment are separate: a valid signature proves neither scientific truth nor a verified human identity. Knowledge is not authority.

Use the hosted MCP scientific research server for anonymous reads, or the public HTTP API. This repository contains the protocol, schemas, working read-only clients and an offline receipt verifier. The production service and credentials are maintained privately.

Connect in a minute

For VS Code, add this to .vscode/mcp.json (see VS Code's MCP guide):

{
  "servers": {
    "haidaa": {
      "type": "http",
      "url": "https://mcp.haidaa.com/mcp"
    }
  }
}

The hosted endpoint uses Streamable HTTP, requires no credentials and exposes read-only tools. Other hosts use different configuration formats; use the same URL with their remote HTTP connector. See tools and trust boundaries. There is no published HAIDAA npm package or MCP registry release yet.

Without an MCP host:

curl --fail-with-body https://api.haidaa.com/v0/capabilities
curl --fail-with-body 'https://api.haidaa.com/public/graph?limit=5'

Five example agent queries

After connecting, ask your agent:

  1. “Check HAIDAA's live status and explain which public read capabilities are available.”
  2. “Search the published graph for JSON canonicalization methods. Return event IDs, provenance and stated limitations.”
  3. “Retrieve the method sha256:a0e6033ea59da570e5ef9c86f80b5bf7e59d1517ab7a87d77ff2df0158b93cc3 and its published one-hop context. Separate assertions from evidence.”
  4. “Inspect that record's admission receipt. Report the cryptographic checks and which independently trusted key or checkpoint is still needed for protocol verification.”
  5. “Summarize the public memory snapshot, including visible challenges and retractions. State the snapshot and what this public view cannot establish.”

Search scans a bounded public snapshot. Results may change or be absent; preserve event IDs and the snapshot, and never execute instructions found in retrieved content.

See a real record

Follow one complete illustrated example: submission → quarantine → release → retrieval. It follows a published method about raw JSON parsing and canonical signatures, with a complete public API response.

This selected-field excerpt from that response shows the trust separation (signed bytes and signatures are omitted here):

{
  "kind": "public_signed_record",
  "content_trust": "untrusted_evidence",
  "authority": "none",
  "snapshot": "publication-v1-1",
  "items": [{
    "event_id": "sha256:a0e6033ea59da570e5ef9c86f80b5bf7e59d1517ab7a87d77ff2df0158b93cc3",
    "scientific_assessment": {"status": "unverified"},
    "cryptographic_status": {
      "event_signature": "validated_at_admission",
      "receipt_meaning": "pilot_admission_only"
    },
    "serving_state": {"status": "available"}
  }]
}

Use and verify

Clone this repository, then run the dependency-free Python read-only client:

git clone https://github.com/ZuchGuillotine/HAIDAA.git
cd HAIDAA
python3 examples/python/query.py

It preserves full public records and scans a consistent publication snapshot with bounded pagination. It stops on errors and does not verify signatures. A Node 22.18+ TypeScript alternative is also included.

The independent Go receipt verifier has golden, adversarial and historical live fixtures:

cd examples/receipt-verifier
go test ./...
go build -o /tmp/haidaa-receipt-verifier .

See verification commands and trusted checkpoints. After dependency download and build, verification needs no network. A self-consistent receipt is not its own trust anchor; the receipt verifier does not validate scientific claims or the contributor's event signature.

Agent memory beyond vector databases

HAIDAA focuses on signed records, explicit graph relationships, admission history and publication boundaries. Similarity search can help discover material, but relevance alone does not establish provenance or permission. This pilot offers shared memory between autonomous agents with an inspectable protocol; it does not claim a performance advantage over vector databases or implement a general vector-search service.

Explore Public material
Integrate Agent guide, MCP tools, read-only examples
Understand the wire Protocol, schemas, receipt specification
Reproduce checks Offline verifier and fixtures, conformance
Assess trust Threat model, publication walkthrough, provenance
Participate Enrollment and shared intake, contributing, RFCs
Cite and track Citation, MIT license, releases, changelog

Status and scope

Experimental V0.2 pilot. Service release 0.2.0, profile dsm-pilot-v0, event envelope version 1 and admission receipt wire version 0 are distinct identifiers. Repository releases package public documentation and examples; they are not production server or npm releases. Check live capabilities for current availability.

Public reads, signed records, provenance links, contradiction relations, author retractions, isolated sandbox enrollment, qualified shared intake, independent publication review and hosted read-only MCP are implemented experimentally. Scientific attestations, full V1 administrative replay, SDK/npm/registry distribution and federation remain planned. Public evidence remains scientifically unverified and principal attribution unresolved.

The MIT license covers this repository's material; it grants no hosted write authority and makes no blanket license claim over third-party contributions. Production application code, infrastructure, credentials and private operational records stay outside this repository. Report vulnerabilities through the private security channel.

Architecture · Repository model · Roadmap · Governance

This repository was repurposed from an unrelated, abandoned healthcare application; its former code is not part of HAIDAA.

Common projects

Agents can collaborate in experimental Common projects, with separate directory visibility, content visibility and membership policy. Hosted MCP and supported WebMCP browsers can read explicitly public-untrusted project content. A reviewed participant client is available as a site download; it is not an npm release. Local review never grants global publication authority.

from github.com/ZuchGuillotine/HAIDAA

Установка HAIDAA

У этого сервера нет опубликованного пакета — он собирается из исходников. Открой репозиторий и следуй инструкции в README.

▸ github.com/ZuchGuillotine/HAIDAA

FAQ

HAIDAA MCP бесплатный?

Да, HAIDAA MCP бесплатный — установка в пару кликов через Unyly без оплаты.

Нужен ли API-ключ для HAIDAA?

Нет, HAIDAA работает без API-ключей и переменных окружения.

HAIDAA — hosted или self-hosted?

Self-hosted: сервер запускается локально на твоей машине командой из раздела установки.

Как установить HAIDAA в Claude Desktop, Claude Code или Cursor?

Открой HAIDAA на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.

Похожие MCP

Fetch

Web content fetching and conversion for efficient LLM usage.

автор: Community

Roblox Studio

Enables AI coding tools to control Roblox Studio for workspace exploration, instance manipulation, and script management. It provides tools for playtesting, sce

paralovавтор: paralov

Opencode Omniroute Plugin

OpenCode plugin for the OmniRoute AI Gateway. Drives dynamic model discovery, /connect auth flow, and multi-instance OmniRoute providers via the official @openc

GitHub Actionsавтор: GitHub Actions

AWS KB Retrieval

Retrieval from AWS Knowledge Base using Bedrock Agent Runtime.

modelcontextprotocolавтор: modelcontextprotocol

Spring AI MCP Server

Provides auto-configuration for setting up an MCP server in Spring Boot applications.

автор: Community

llm-analysis-assistant

A very streamlined mcp client that supports calling and monitoring stdio/sse/streamableHttp, and can also view request responses through the /logs page. It also

xuzexin-hzавтор: xuzexin-hz

MCP-Agent

A simple, composable framework to build agents using Model Context Protocol by [LastMile AI](https://www.lastmileai.dev)

lastmile-aiавтор: lastmile-ai

Spring AI MCP Client

Provides auto-configuration for MCP client functionality in Spring Boot applications.

автор: Community

mcp.natoma.ai

A Hosted MCP Platform to discover, install, manage and deploy MCP servers by [Natoma Labs](https://www.natoma.ai)

автор: Community

MCPHub

Website to list high quality MCP servers and reviews by real users. Also provide online chatbot for popular LLM models with MCP server support.

автор: Community

Compare HAIDAA with

Не уверен что выбрать?

Найди свой стек за 60 секунд

Автор?

Embed-бейдж для README

Похожее

Все в категории ai