MCPJSON-RPC 2.0Groq

MCP Tool Demo

Model Context Protocol — real-time tool discovery, selection, and execution

MCP standardizes how AI agents discover and call tools. Watch the full JSON-RPC 2.0 lifecycle: tool schema negotiation, LLM selection, server-side execution, and answer synthesis — all transparent and traceable.

Tools Available

get_experience · search_skills · get_achievements

Transport

HTTP

JSON-RPC 2.0 over HTTPS

Standard

MCP

Linux Foundation · open spec

Protocol Lifecycle

Discover

Client → Server: tools/list request

Select

LLM selects tool(s) from schema

Execute

Server runs tool · returns result

Synthesize

LLM forms final answer from results

Tool Registry

get_experience

Retrieves work history, roles, and tenure context

(company: string) → ExperienceRecord

access: public

search_skills

Returns skills for a category

(category: string) → SkillRecord[]

access: public

get_achievements

Returns quantified business outcomes — requires an agent credential

(company?: string) → Achievement[]

access: scope: read:profile

Per-tool authorization is enforced on the server with default deny. Ask for achievements without a credential to see a denied call in the log; the Agent Auth demo issues a read:profile credential that unlocks it. Tool outputs are screened for injected instructions before the model sees them.

Run a Query

Ctrl+Enter to run

Why MCP matters for enterprise AI

Standardization

One protocol for all tool integrations — no bespoke adapters per agent or model provider.

Auditability

Every tool call is a traceable JSON-RPC message with explicit input/output — no black-box side effects.

Governance

Tool schemas enforce capability contracts. Rate limiting and guardrails layer on top of the protocol.

Browse all MCP-enabled demos in the Agent Marketplace →