"Keep a Changelog" matching MCP servers:

Ships only the MCP server, not the recorder. Recording works by wrapping a process on the host — a local proxy, a PATH shim, a shadow git index. None of that survives a container boundary, so orca record inside here would watch an empty namespace. What does travel is the read side: the six MCP tools serve runs that already exist on disk, so a client mounts its .orca directory and r

Enables users to log AI tool activity without conversation text and contribute back only user-approved, small-cell-suppressed counts through a local MCP interface.

Tamper-evident audit logging for AI agents. Append-only, hash-chained, optionally Ed25519-signed log. The MCP server lets an agent keep and verify a record of what it actually did.

Enables read-only querying and analysis of Kibana logs through a logged-in sid Cookie, including keyword/time/index searches, traceId lookups, surrounding log context, and aggregations by service, level, or host. It also validates Kibana connection status and redacts sensitive fields such as cookies, tokens, and passwords.

An MCP server for querying Grafana Loki directly with a discovery-first workflow — labels, values, series, and LogQL queries without requiring Grafana.

An MCP server that records agent execution metrics and exposes a Context Window Explorer to visualize exactly what entered the model's context window across sessions, tokens, and tool calls.

Provides cryptographically signed, tamper-evident compute usage receipts via a Model Context Protocol server, exposing tools to record usage, verify receipts, list the ledger, and verify ledger integrity without requiring any account, API key, or single cloud provider.

Tells an agent what went wrong with its own Ollama calls, in sentences with the fix: lost prompt cache and where the prompt stopped matching, context overflow, a model running partly on CPU, tool schemas crowding the prompt. Also the last turn in numbers and which local models fit in video memory. Local, read-only, nothing leaves the machine.

8 standalone MCP servers for SMBs — Guardrails (prompt injection, PII, secrets detection), FinOps (AI cost tracking and budgets), Observability (agent tracing), Smart Router (17+ models, 9 providers), Trust Score (governance scoring 0-100), Memory (persistent agent memory), ThinkSecure (runtime security + SOC2/GDPR audit), and A2A Lite (agent-to-agent protocol). 37 tools total, TNC credits

MCP server that diagnoses Apache Spark job failures and optimizes performance using stack-trace analysis and LLM providers, supporting EMR and local sources.

A read-only MCP server that exposes local coding-agent session logs as three tools for introspection of recent work, debugging tool failures, and tracking token usage and estimated cost without parsing log files.

Compares AI coding tool spend before and after a migration -- Cursor, Claude Code, GitHub Copilot, OpenCode, Codex CLI -- via an MCP server, real numbers from each vendor's own API or local logs.

Enables analysis of JVM thread dumps to detect deadlocks, CPU spikes, pool exhaustion, and virtual thread pinning, returning a health score and interactive report link.

Provides a local MCP interface to query captured, PII-scrubbed prompt telemetry from coding agents and browser chats, enabling time summaries, prompt search, and activity analysis without data leaving the machine.

Enables agents to interact with the Logflare API by writing JavaScript functions that run in a Vercel Sandbox, using three tools: search, execute_read, and execute_write.

An MCP server that connects language models to a Wazuh SIEM cluster, enabling read-only plain-language queries, schema inspection, field coverage analysis, and decoder testing.

Enables natural language queries of the Chrome UX Report API to fetch Core Web Vitals (LCP, INP, CLS, FCP, TTFB), compare form factors, benchmark pages against site averages, and analyze up to 40 weeks of performance trends.

An MCP server that exposes a dbt project's run state as tools, enabling AI assistants to answer questions like 'is the warehouse healthy?' or 'what broke and why?' via plain English, using read-only access to dbt artifacts and warehouse data.

Enables an LLM agent to investigate a Kong Gateway through fourteen read-only, schema-validated tools and a deterministic diagnostic engine, so it can trace routes, services, plugins and upstream health and explain faults such as why authentication is failing on a path. Every fact comes from the gateway itself, and the model only selects tools and writes the evidence-backed explanation.

Model Context Protocol server for Verlon AI that exposes gates, logs, recommendations, and experiments as MCP tools, enabling coding agents to inspect and manage AI infrastructure natively.