MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
AI work receipts and deterministic verification for agents, with MCP, evidence, claims, approvals, and SHA-256 integrity.
MIT -
Local-first MCP server for literature, Zotero, writing, and versioned research evidence with bounded responses.
MIT ★ 1 -
evidence-enrichment-enginegithub.com/tinnlo/evidence-enrichment-engineupdated this year
Replay-first evidence enrichment pipeline with hierarchical retrieval, schema-validated typed extraction, AI FinOps cost governance, execution policy capability control (off|audit|enforce), Langfuse-first observability, optional LangGraph retrieval, and MCP tooling.
-
Private, evidence-first document agent with persistent local memory, exact citations, reversible actions, and bounded Nemotron reasoning.
MIT ★ 1 -
research-connector-hubgithub.com/nicks2025fk/research-connector-hubupdated this year
Local-first research evidence and structured brief MVP with CLI and a bounded MCP adapter.
MIT -
Free-first Chinese and English web search MCP using zero-key sources and inspectable evidence.
npm Apache-2.0 ★ 113 -
Local-first, Agent-callable Evidence engine for documents and media, exposed through Python, CLI, and MCP.
MIT -
A tiny local-first raw evidence ledger for AI and human workflows.
mcpb MIT ★ 1 -
Offline-first scam triage and evidence kit: explainable verdicts, scam playbooks, indicator extraction and tamper-evident cases. Stdlib Python, no dependencies, MCP server included.
MIT -
MCP-first review and evidence-mapping harness for AI-generated and human-drafted deliverables.
MIT ★ 1 -
Open-source, local-first investor tooling where every number shows its source. Evidence ledger, claims audit, MCP server. Information, not advice.
Apache-2.0 -
Local-first research tool with explicit fetch decisions, source evidence, and cited reports. CLI and MCP interfaces. Python, MIT.
pypi MIT ★ 61 -
Evidence-driven RAG. Hybrid BM25 + vector retrieval with citations you can re-verify. Local-first, zero API keys, MCP-native.
Apache-2.0 -
Local-first project management for Codex: tasks, dependencies, evidence, human approvals, chat history, CLI and MCP.
MIT -
Personal, local-first conversation memory and evidence retrieval for AI coding assistants, with provenance, evaluation, and recovery.
MIT -
Codex-native-long-video-memorygithub.com/NitcanKen/Codex-native-long-video-memoryupdated this year
Local-first Codex plugin that indexes long videos with OCR/ASR and answers with timestamped, SHA-verified visual evidence.
Apache-2.0 ★ 4 -
Evidence-first cybersecurity MCP server with persistent SQLite graph memory, vulnerability triage, scoped testing and reports for Antigravity and Codex.
MIT ★ 1 -
Local-first MCP server, CLI and SDK for evidence, formal checks, planning and experimental image-to-3D workflows. Includes real mesh demonstrations, reproducible comparisons and model provenance. Developer alpha.
npm MIT -
Evidence-first Counter-Strike 2 demo analysis and replay review. Local JSON CLI + MCP server for player stats, round context and AI-agent workflows.
-
Local-first Android reporter that turns bug reports into portable, reviewed evidence parcels for humans and coding agents. Offline validator, viewer and read-only MCP. AGPL-3.0-only; source candidate, not yet released.
Apache-2.0 -
Loop Factory MCP — a local-first supervisor for evidence-gated AI agent loops. Mines past agent sessions for workflows that worked, improves them generation over generation, and only promotes a measured, re-verified win. Zero dependencies, MCP over stdio.
MIT -
Open-source, evidence-provenance-first scientific literature MCP server: 35 tools for literature search, citation/retraction checks, full-text quote verification, methodology audits, PRISMA-style screening, hash-chained evidence ledger, BM25 retrieval and NLI assessment. Apache-2.0.
Apache-2.0 ★ 5 -
Recordings in. Agent-ready evidence out. Local-first MCP for transcripts, keyframes, OCR, speakers, and wall-clock evidence.
pypi MIT ★ 34 -
Run unit tests on your manuscript. Check every manuscript claim against its cited evidence. Human-first, AI-second, Zotero-audited, local-first.
pypi Apache-2.0 ★ 1 -
@sylphx/image-reader-mcpgithub.com/sylphxai/image-reader-mcpupdated this year
Iris — Evidence-first image reading for AI agents — metadata, OCR text, regions, and citeable evidence without generative LLM.
npm MIT -
Powerful local-first browser automation for AI agents with real Chromium, isolated profiles, MCP, Maqam approvals, and verifiable evidence.
npm AGPL-3.0 ★ 9 -
Evidence-first research CLI and MCP server with replayable source-linked reports. Install via uv from GitHub.
MIT ★ 1 -
Local-first Taiwan market evidence workbench for AI agents — TWSE/TPEx, persistent Watchlists, governed MCP, auditable evidence, explicit execute-once.
mcpb MIT -
@pcircle/footprintgithub.com/pcircle-ai/footprint-mcpupdated this year
Local-first MCP server for AI work history, context memory, and encrypted evidence
npm MIT -
Evidence-first CLI and optional MCP server for timestamped video transcript search and citation validation.
MIT ★ 1 -
Harness-neutral, evidence-first company research with versioned contracts, MCP adapters, and completed-only publication.
Apache-2.0 ★ 1 -
Evidence-first verification layer for delegated AI work. Do not trust completion claims — verify the work.
npm MIT -
research-navigator-agentgithub.com/maryam-na/research-navigator-agentupdated this year
Local-first research discovery agent with ADK-facing orchestration, MCP-ready tools, evidence grounding, and Streamlit review dashboard.
-
ProofGate Runtime v0.2 - evidence-first claim verification (confirmed/refuted/unverified), HMAC-signed receipts, MCP server. Stdlib only; not independently audited.
pypi Apache-2.0 -
@sylphx/video-reader-mcpgithub.com/sylphxai/video-reader-mcpupdated this year
Cue — Evidence-first video reading for AI agents — ffprobe, subtitles, scenes, transcripts, and timelines without frame-by-frame LLM vision.
npm MIT -
Local-first, evidence-backed research memory and writing workspace
MIT -
Local-first, evidence-backed memory daemon for MCP-capable AI Agents
MIT -
Zero-dependency security auditor for Dockerfiles, compose files and .dockerignore — three passes, evidence-first, SARIF for CI.
MIT -
Evidence-first financial research — every signal traced to its SEC filings and verifiable against a public, git-anchored ledger. CLI + REST + MCP + LangChain/LlamaIndex. Research / education only — not investment advice.
pypi Apache-2.0 -
Review AI coding deliveries with evidence, exact commits, and human decisions. Local-first kanban + MCP.
MIT ★ 1 -
Bonsai-backed local evidence and memory layer for frontier coding agents. MCP-first with an OpenCode plugin.
MIT ★ 2 -
Evidence-first reverse engineering workbench with bounded binary analysis, extensible static adapters, MCP and reproducible local investigations.
MIT -
Evidence-first Android and iOS security audits with CLI, TUI, MCP, public vulnerability intelligence, and OWASP-aligned evidence.
pypi MIT -
Community ReMe companion for evidence-first current facts and tasks | Markdown + CLI + MCP
Apache-2.0 -
rescicleactive
Local-first research copilot: MCP server and Research View for questions, hypotheses, predictions, and evidence (RRS)
npm Apache-2.0 -
Local-first, AI-native desktop workbench for product decisions: evidence, competitors, exact unit economics, and an evidence-linked report in one project folder you own. AI drafts and waits; nothing acts on its own. Read-only MCP server included. No cloud, no accounts, no telemetry.
npm AGPL-3.0 ★ 7 -
Evidence-first MCP timeline debugger with reversible provenance replay, Supported/Unknown/Conflict states, and auditable intervention receipts.
MIT -
Collector-first engineering evidence backend for deterministic ingestion, Qdrant retrieval, cited investigation, readiness reports, workflows, evals, and Core MCP. Azure deployment scaffold; not yet production-grade.
★ 1 -
Evidence-first local MCP and skills for turning AI research into measured, human-approved setup changes.
npm Apache-2.0 -
Real-time interactive development notes & evidence-backed project control center for solo builders and teams coding with AI (MCP + Local-First)
Apache-2.0 -
A local-first project spine for coding agents: shared context, durable tasks, verification evidence, and seamless handoffs.
MIT -
Local-first codebase discovery for AI agents, with precise navigation and verifiable evidence.
Apache-2.0 -
BingguPack — local-first, evidence-backed memory/context pack framework with an installable Claude Code MCP server (stdio JSON-RPC). Python stdlib only.
pypi MIT -
SkillRoute by JEStats — local-first skill routing for coding agents, with ranked matches, confidence, and evidence over MCP.
npm MIT ★ 41 -
Voice-first paperwork assistance for blind and low-vision users: MCP Apps, user-defined action limits, and a tamper-evident ledger.
MIT -
Trace reviewer feedback to the evidence that it was addressed in a revised document. Local-first, evidence-based DOCX review auditor with CLI and MCP server.
MIT -
Verification-first mobile automation for iOS and Android: typed plans, fresh observations, evidence-backed results, and an MCP server. Rust engine, local-first.
Apache-2.0 -
Local-first workflow compiler that turns repeated AI-agent work into evidence-backed, quality-checked Skills.
MIT ★ 2 -
A local-first memory layer for AI tools. One SQLite file, evidence receipts, and a review queue. Built with MCP.
npm MIT -
Miragithub.com/zxheyi/Miraactive
Local-first project memory for coding agents. Preserve context across sessions with auditable memories, MCP integration, and evidence-gated research.