MCP Tools for AI Agents

A working AI agent is mostly tools. The model decides; the tools do. agentFast ships 50+ MCP-native tools out of the box and a wrapper pattern that makes adding your own — or any Model Context Protocol server — a matter of config, not code.

50+ tools behind one registry
MCP tools for AI agents: 50+ built in, add any MCP server by configTools are grouped by agent type behind one @tool wrapper and a fixed 12-code error vocabulary. Any MCP server added by config is auto-namespaced, traced, rate-limited and HITL-overridable.core5 toolsshared11 toolssupport13 toolsresearch13 toolscode12 toolstool registry — one @tool wrapper · fixed 12-code error vocabularyyour MCP servers (by config)auto-namespacedtracedrate-limitedHITL-overridableINVALID_INPUT · TIMEOUT · INTERNAL_ERROR · NOT_FOUND · RATE_LIMITED · … (12 codes, never free-text)
What makes them production tools

50+ tools, grouped by the work

Core, shared, support, research and code groups: web/KB search, read-only-guarded SQL, a sandboxed shell, SSRF-guarded fetch, ticketing, real arXiv/Semantic Scholar APIs.

One error vocabulary

Every tool speaks the same fixed 12-code vocabulary — INVALID_INPUT, TIMEOUT, INTERNAL_ERROR — so the model gets structured, predictable failures it can reason about.

One @tool wrapper

Validation failures become INVALID_INPUT, timeouts become TIMEOUT, crashes become INTERNAL_ERROR — enforced in one place.

Add any MCP server by config

Point agentFast at a server and its tools are discovered, auto-namespaced, traced and rate-limited like any built-in — no orchestration code.

Add a server in a few lines
agentfast.yaml
mcp_servers:
  - name: billing
    transport: stdio
    command: node
    args: ["./mcp/billing-server.js"]
Guardrailed and observable by default
Every tool call flows through guardrails and lands in the step tree with its own cost and latency — and you can re-classify any tool as high-risk so it routes through human approval.

For the deeper argument on tooling, read the 50 MCP tools every production agent needs.

Stop rebuilding the plumbing.

agentFast is the production layer — memory, 50 tools, observability, evals, guardrails, human-in-the-loop — for LangGraph, CrewAI, the Claude & OpenAI Agent SDKs & Vanilla. Own it for life.

Get agentFast — from $89
Keep reading
The Production Layer Every AI Agent Needs (and Why LangGraph Doesn't Give It to You)
9 min read
LangGraph vs CrewAI vs a Vanilla Loop: Which Should You Actually Use in 2026
11 min read
How to Add Human-in-the-Loop Approval to Any AI Agent (With Durable Execution)
10 min read