Neonia

A collection of autonomous agent examples demonstrating how to integrate theNeonia Model Context Protocol (MCP) Gatewayusing the officialStreamable HTTPtransport standard.

This repository will continuously grow with new patterns demonstrating deterministic, high-performance AI agents.

Our first major showcases focus on solving two critical problems in modern agent architectures:Context Window BloatandTool Rigidity.

1. Solving Tool Rigidity (Auto-Pilot Discovery)

Agents are traditionally hard-coded with a static list of tools. If a user asks for something outside that list, the agent hallucinates or fails. Theauto-discovery-url-to-markdownexamples demonstrate how to give your agents true autonomy. By connecting to the Neonia Gateway, the agent can dynamically search for missing capabilities, read the tool's schema, and execute it on the fly without human intervention.

2. Solving Context Bloat (Zero-Bloat Data Processing)

Traditionally, when an agent needs to extract data from a large 5MB JSON file, it loads the entire file into its context window, causing massive token consumption, high latency, and LLM "amnesia". By connecting to the Neonia MCP Gateway (mcp.neonia.io/mcp?tools=neonia_data_jq_filter), ourzero-bloat-jq-filteragents explicitly bind theWasm-powered JQ Filtertool. The agent executes queries on the remote server and receives only the filtered result (e.g.$651,758.23), saving~50,000+ tokensper request and responding almost instantly.

3. Stateful Memory Note (stateful-cloud-memory)

Agents typically suffer from absolute amnesia between sessions. If a user states a preference or business rule, it is lost unless hardcoded into the system prompt. Thestateful-cloud-memoryexamples demonstrate how to create stateful agents that use Neonia's Dual Memory Architecture (neonia_sys_memory_notefor writing andneonia_sys_memory_searchfor reading) to dynamically store and recall rules (like custom personas or user preferences) across completely isolated sessions without needing a custom database.

4. Persistent Knowledge Memory (persistent-knowledge-memory)

When agents learn hard architectural lessons, bug fixes, or strict operational rules, they need a way to persist this knowledge using a strict Cause-and-Effect structure (ADR). Thepersistent-knowledge-memoryexamples demonstrate using theneonia_sys_memory_lessontool to save complex insights so future agents can fetch them usingneonia_sys_memory_searchbefore starting their tasks.

(More examples covering vision extraction, dynamic execution, and multi-agent orchestration will be added soon!)

This repository includes implementations of "Zero-Bloat Data Processing", "Auto-Pilot Tool Discovery", and "Stateful Memory Note" across major agentic frameworks in 3 different languages:

A deterministic workflow usingLangChainandLangGraphto build a reactive agent (create_agent) that dynamically wraps MCP capabilities into native LangChain@toolinstances.

- Directories:python/langgraph/zero-bloat-jq-filter,python/langgraph/chained-json-jq-filter,python/langgraph/auto-discovery-url-to-markdown,python/langgraph/stateful-cloud-memory,python/langgraph/persistent-knowledge-memory
- Setup:uv sync && uv run python agent.py

A self-assembling agent using Hugging Face'sSmolAgentsandLiteLLM. Demonstrates subclassingsmolagents.Toolfor synchronous forward execution wrapped around an asynchronous Streamable HTTP session.

- Directories:python/smolagents/zero-bloat-jq-filter,python/smolagents/chained-json-jq-filter,python/smolagents/auto-discovery-url-to-markdown,python/smolagents/stateful-cloud-memory,python/smolagents/persistent-knowledge-memory
- Setup:uv sync && uv run python main.py

An integration with theVercel AI SDKutilizing the official@modelcontextprotocol/sdkand@openrouter/ai-sdk-provider. Demonstrates proper multi-turn tool calling and schema mapping for Claude 3.7 Sonnet.

- Directories:typescript/vercel-ai-sdk/zero-bloat-jq-filter,typescript/vercel-ai-sdk/chained-json-jq-filter,typescript/vercel-ai-sdk/auto-discovery-url-to-markdown,typescript/vercel-ai-sdk/stateful-cloud-memory,typescript/vercel-ai-sdk/persistent-knowledge-memory
- Setup:npm install && npm start

A statically-typed integration using theRigagent framework andrust-mcp-sdk. Demonstrates bridging an initialized MCP client session into Rust's strong type system.

- Directories:rust/rig/zero-bloat-jq-filter,rust/rig/chained-json-jq-filter,rust/rig/auto-discovery-url-to-markdown,rust/rig/stateful-cloud-memory,rust/rig/persistent-knowledge-memory
- Setup:cargo run
- ANeonia API Key(NEONIA_API_KEY)
- AnOpenRouter API Key(OPENROUTER_API_KEY)

Configure these in the.envfile within the specific example directory you wish to run.

agent-mcp-examples/ ├── typescript/ # TypeScript Ecosystem │ └── vercel-ai-sdk/ # Vercel AI SDK Framework │ ├── zero-bloat-jq-filter/ # Single-tool Data Processing │ ├── chained-json-jq-filter/ # Multi-tool Chained Data Processing │ ├── auto-discovery-url-to-markdown/ # Auto-Pilot Tool Discovery │ ├── stateful-cloud-memory/ # System Memory Note Persistence │ └── persistent-knowledge-memory/ # Architectural Lesson Persistence │ ├── python/ # Python Ecosystem │ ├── langgraph/ # LangGraph Framework │ │ ├── zero-bloat-jq-filter/ │ │ ├── chained-json-jq-filter/ │ │ ├── auto-discovery-url-to-markdown/ │ │ ├── stateful-cloud-memory/ │ │ └── persistent-knowledge-memory/ │ └── smolagents/ # SmolAgents Framework │ ├── zero-bloat-jq-filter/ │ ├── chained-json-jq-filter/ │ ├── auto-discovery-url-to-markdown/ │ ├── stateful-cloud-memory/ │ └── persistent-knowledge-memory/ │ └── rust/ # Rust Ecosystem └── rig/ # Rig Framework ├── zero-bloat-jq-filter/ ├── chained-json-jq-filter/ ├── auto-discovery-url-to-markdown/ ├── stateful-cloud-memory/ └── persistent-knowledge-memory/

Each example is self-contained and demonstrates specific, production-ready architectural patterns over MCP.

1. Auto-Pilot Tool Discovery (auto-discovery-url-to-markdown)

Demonstrates how to give agents true autonomy. If an agent lacks a required capability, it dynamically searches the Neonia Gateway for a matching tool, reads its parameters, and executes it on the fly without human intervention.

- 📂auto-discovery-url-to-markdown(TypeScript / Vercel AI SDK)
- 📂auto-discovery-url-to-markdown(Python / LangGraph)
- 📂auto-discovery-url-to-markdown(Python / SmolAgents)
- 📂auto-discovery-url-to-markdown(Rust / Rig)

- 📂zero-bloat-jq-filter(TypeScript / Vercel AI SDK)
- 📂zero-bloat-jq-filter(Python / LangGraph)
- 📂zero-bloat-jq-filter(Python / SmolAgents)
- 📂[zero-bloat-jq-filter(Rust / Rig)

3. Chained Data Execution (chained-json-jq-filter)

…