"Information about costs or pricing" matching MCP servers:

MCP-native agent evaluation and observability server. Log traces, evaluate output quality with 12 built-in rules (PII detection, prompt injection, cost thresholds), and track agent costs. Real-time dashboard, OTel-compatible spans. Self-hosted, MIT licensed.

Evaluates task suitability for local models before cloud API calls, routing to Ollama or similar to reduce costs.

A reference implementation of the mcp-lens pattern for progressive disclosure in MCP servers. It exposes three stable meta-tools to search, inspect, and execute capabilities, keeping tool-definition costs constant as the catalog grows.

Lets AI agents rent GPUs on Nosana, browse deployable templates, compare live GPU prices, estimate costs, and deploy or manage services like MiniMax H3 using account credits.

Just to be clear about the final JSON: the MCP server is named herdr-inbox, and it enables Claude Code agents in herdr to ask questions asynchronously without stalling, with all open questions answered from one popup and answers routed back to the asking agent.

MCP server that scores tool descriptions, estimates token costs, simulates agent tool selection, and generates reliability reports to help AI agents choose the right tools and reduce wasted tokens.

Provides AI assistants with real-time access to the AGNO framework documentation by enabling them to browse, search, and fetch documentation pages. This server allows users to query information about AGNO's Agents, Teams, and Workflows directly through MCP-compatible clients.

Connects AI assistants to Warpmetrics telemetry data to monitor AI agent performance, execution runs, and LLM costs. It allows users to query success rates, latency, and spend metrics directly through natural language interfaces.

Classifies AI agent turns by complexity and recommends the cheapest model that can handle them, reducing costs in multi-step workflows.

Enables selecting a saved Codex or Claude Code session from the same PC and importing its recent context into the other tool, so work can continue across clients with source and omission information.

A Model Context Protocol server that enables AI assistants to interact with Hedera blockchain in real-time, allowing them to check account balances, view block information, and estimate transaction costs.

Enables agents and users to route prompts to the cheapest reachable free or paid model across providers, compare costs, list filtered models, and receive fallback chains and cached routing results without calling any LLM API.

Compute price oracle for AI agents. Compare inference pricing across OpenAI, Anthropic, and DePIN providers like Hyperbolic. Get routing recommendations that save up to 80% on compute costs.

Enables agents to crawl and read websites politely, learning per host which reader, URL rewrite, or detail path actually reaches good information, and reusing those wins across runs or a shared database. Failures come back as coded errors with concrete next steps, so blocked or empty pages point to an honest alternative instead of endless retries.

A persistent memory implementation using a local knowledge graph that lets Claude remember information about users across conversations.

The AgentOps MCP server provides access to observability and tracing data for debugging complex AI agent runs. This adds crucial context about where the AI agent succeeds or fails.

Enables Claude Code and Gemini CLI to see your macOS screen, answering questions about the file, error, UI, database table, or web page you're looking at. Captures screenshots, reads database schemas, and controls Chrome/Brave, presenting answers as on-screen highlights or spoken bubbles above your cursor.

A Model Context Protocol server that enables Claude Desktop to play the classic Lemonade Stand business simulation game with dynamic weather, pricing strategies, and inventory management.

Cuts AI token costs by running user code in a secure sandbox, so data never enters the context window, enabling efficient data processing with MCP protocol.

Compresses multi-turn conversation history by stripping verbose boilerplate and filler tokens while preserving critical tool parameters and named entities, deterministically cutting inference costs by 40-60%. Runs as a zero-dependency Python MCP server or importable module, compatible with Claude Desktop, Cursor, and other MCP runtimes.

Drive Clearcote, a Chromium build with its fingerprint protection patched into the engine, from any MCP client: about 20 tools to navigate, read, extract, click, fill, screenshot and keep sessions, plus a CDP endpoint that hands the same browser to Playwright or Puppeteer. The server lives in the repo's mcp/ folder (npx clearcote-mcp or pip install clearcote-mcp).

A CLI and cloud platform providing managed hosting for MCP servers with intelligent tool filtering to reduce token costs and improve LLM accuracy. It serves as a secure proxy that connects AI clients to cloud tools while keeping sensitive credentials on the local machine.

MCP server that defers non-urgent tasks to off-peak hours, saving ~50% on DeepSeek API costs without blocking users.

Enables AI assistants to manage Paperclip AI agents, issues, goals, approvals, and costs via natural language by exposing Paperclip's REST API as MCP tools.

Enables AI agents to pre-flight estimate query costs, enforce hard budgets, reconcile actual billed spend, and suggest cheaper rewrites across BigQuery, Snowflake, and Databricks.

Enables selecting the best AI model and reasoning effort for any task by analyzing task requirements and live pricing, balancing intelligence, speed, and cost. Provides model recommendations, comparisons, and routing status through MCP tools.

Enables supported AI clients to discover, inspect, and execute 2,000+ models and APIs, including free data APIs and paid LLM, image, video, and audio calls, with OAuth, pricing checks, and rollback.

Discovers LLM models in real time from cloud providers and local Ollama instances, returning compatibility profiles and live pricing so AI agents can route tasks to the cheapest viable model without breaking tool calls or context clipping.

Keeps large tool outputs out of an agent's context window by spilling them to disk and handing back a short pointer instead. Lets the agent pull only what it needs — exact numbered line ranges read directly from disk, or verbatim relevant lines returned by a small model answering a question about the stored output, plus tools to open, list, and close documents.

Enables an LLM to drive a real browser by declaring one outcome at a time, while Jev chooses the element, action, and value and Playwright executes it, returning a clear status such as done, blocked, or needs confirmation. Also exposes tools to open pages, ask yes/no or multiple-choice questions about a page, read question-relevant page text, capture snapshots and screenshots, act on elements directly, and invoke a page's own WebMCP tools.