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.
-
Failure memory and epistemic guardrails for AI agents
pypi MIT -
Local-first agent memory for research: literature evidence, experiment runs, and failure cases with provenance contracts, quantified effectiveness metrics, and a read-only MCP server for Codex, Claude Code, and Cursor.
-
@bambushu/scargithub.com/bambushu/scarupdated this year
Failure memory for AI agents - briefing hooks + optional MCP server
npm MIT -
jambavangithub.com/beingmartinbmc/jambavanupdated this year
Local-first MCP memory and code intelligence that helps coding agents recall repo context, decisions, failures, and review impact.
npm MIT -
Local-first MCP server that gives any coding agent a persistent memory of your repo: AST code index, durable decisions, failure ledger, review packs. No LLM calls, no telemetry, no code upload.
MIT ★ 2 -
Temporal Knowledge Graph System for Architectural Memory - Track decisions, patterns, and failures with MCP integration for Claude Code/Desktop
MIT ★ 4 -
ejentum-mcpno update data
Exposes the four Ejentum cognitive harnesses (reasoning, code, anti-deception, memory) as MCP tools any agentic client can call. Drop-in scaffolding that catches LLM failure modes like sycophancy, hallucination, and reasoning shortcuts.
smithery -
agentedgithub.com/frane/agentedno update data
A text editor for LLMs, not humans. Take ed, the line editor that nobody has voluntarily used since about 1975, and rebuild it for an environment where the typing user is a language model. Short verbs, line addresses, no modes, no TUI. What an editor optimises for changes when the user is the model: round trips per task, tokens per command, an editing buffer with a long memory, and an undo tree that remembers the branches the agent abandoned, because that's often where the interesting work was. ## Features - **Fewer round trips.** Read-before-Edit is unnecessary; on conflict the response carries the new content so you reconcile in one call instead of pre-reading every time. - **Branching undo.** Walked-back work stays addressable instead of being thrown away when you pick a different path. - **Three-way merge.** Concurrent agents get a structured conflict response instead of a silent overwrite. - **Atomic batches.** Multi-file refactors run all-or-nothing instead of leaving half-applied state on failure. - **Cross-file moves and regex replace as primitives.** Operations the built-in tools can't express cleanly become single calls. - **Drift detection.** External edits to an open file are folded into the tree instead of being clobbered by the next write. - **Inline diagnostics.** Type errors and lint findings surface on save, not at the next build many edits later. - **Cross-session memory.** Per-file notes persist between sessions and surface inline on the next open. - **Audit log.** Every operation recorded with actor and timestamp, so two agents in one workspace can't argue about who moved the head. ## Multi-workspace (v0.4.0) One MCP server registration handles every project you mention. Tool calls route to the workspace owning the absolute path you pass; each project gets its own SQLite, its own undo tree, its own LSP daemon — lazily spawned on first touch. Desktop hosts (Claude Desktop, Codex Desktop, Cursor) install the server once globally and use it across projects without per-project config. ## Install `ae` must be installed locally; this MCP bundle registers `ae serve` as a stdio server. ```sh brew tap frane/tap && brew install agented # or curl -sSL https://raw.githubusercontent.com/frane/agented/master/install.sh | sh ``` Pure Go, single static binary, Apache 2.0. Source: https://github.com/frane/agented
smithery -
Re-debugging Solved Failures? No Way. 1000+ Evidence-graded fix lessons, One MCP Call-->https://misakanet.org
npm Apache-2.0 ★ 561 -
kira-mcpgithub.com/aibenyclaude-coder/kiraupdated this year
Where agents shine — auto-manages Skills and Scars for AI agents via MCP.
npm MIT -
Persistent codebase memory for AI coding agents — Claude Code, Codex, Cursor, Cline, Continue and any MCP client. Learns which files go together from how you work and serves compact, ranked context: the gold file on 40 of 40 public-benchmark queries to keyword ripgrep's 32 of 40, at 31-67× fewer tokens than ripgrep and 51-242× fewer than pasting every source file (a larger 150-question eval misses 10). Local-first, no telemetry by default.
pypi MIT AND LicenseRef-NeuralMind-Commercial -
Shared, append-only context store for AI agents over MCP: checkpoints, failures, claims, handoffs, crash recovery. Multi-process safe.
npm MIT -
ThumbGate Pre-Action Checks self-improve from ranked lessons and repeated failures, hard-block detected secret leaks, and block matches in strict mode.
npm MIT -
common context MCP — shared, append-only context store for AI agents over MCP: checkpoints, failures, file claims, handoffs and crash recovery. npm: https://www.npmjs.com/package/common-context-mcp
MIT ★ 3 -
Cognitive reconstruction engine for AI agents — Python SDK + CLI. Loop-Ready Contract: cursor-paged settle feed (snapshot_transitions: next_cursor, horizon_tx, truncated) + structural causal model view (snapshot_scm); KARM contract surface: CognitiveSnapshot with laws/policies/failures/uncertainty, CognitiveStateProvider/Sink traits; LawExtractor; PolicyRegistry; MDL sparsity GC; 64-tool / 7-resource MCP; WAL-persistent. LangChain, AutoGen, CrewAI, MCP
pypi Apache-2.0