Hyperterse – Agentic Server Framework for MCP and A2A

TL;DR

TL;DR: Hyperterse is an open-source agentic server framework that bundles A2A agents, MCP tools, prompts, and database adapters into one compile-and-deploy process — replacing a stack of separate services with a single config-driven binary.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: hyperterse.com — verified, returned HTTP 200
• Source repository: github.com/hyperterse/hyperterse — README read in full
• License: Apache-2.0 (verified via GitHub API license.spdx_id)
• HN launch thread: news.ycombinator.com/item?id=47142975 (Show HN, 2026)
• Source last checked: 2026-07-08

What Is Hyperterse?

Hyperterse is an agentic server framework: one build ships agents (A2A), tools (MCP), prompts, resources, database adapters, auth, caching, and observability from a single process. You declare surfaces in config files; the compiler validates and bundles them. At runtime, clients connect via MCP Streamable HTTP at /mcp for tools, prompts, and resources, and via A2A-style agent routes at /agent/{name}.

The README describes it as:

Hyperterse is an agentic server framework: one build ships agents (A2A), tools (MCP), prompts, resources, database adapters, auth, caching, and observability from a single process.

The core idea is that you no longer need separate services for agent orchestration, tool exposure, and data access. Instead, you define what your agents can do and what data they can reach in declarative config, and Hyperterse compiles it into one deployable artifact.

Core Capabilities

Agents (A2A): Define declarative agent configs with tool-access policies, multi-provider model support, and per-agent A2A HTTP routes.

MCP Tools: Each tool definition becomes one MCP endpoint. Supports database-backed tools (use + SQL statement) and script-backed tools (TypeScript handler). File-system discovery means the compiler finds and bundles all tools automatically.

Database Adapters: Built-in adapters for PostgreSQL, MySQL, SQLite, MongoDB, and Redis — no extra glue code needed.

Prompts and Resources: Reusable prompt templates and static context surfaces, following the same discover-and-compile model as tools.

Execution Pipeline: Every tool call runs through resolution, authentication, optional input transform, then DB or handler execution.

Auth and Caching: Built-in allow_all and api_key auth strategies, plus custom plugins. In-memory caching with global defaults and per-tool overrides.

Observability: OpenTelemetry tracing and metrics, structured logging out of the box.

Setup Workflow

Step 1: Install

curl -fsSL https://hyperterse.com/install | bash

Step 2: Scaffold a project

npx create-hyperterse-app my-agentic-service
cd my-agentic-service

Step 3: Define your surfaces

Create tools/, agents/, prompts/, and resources/ directories with declarative config files. The compiler validates all references at build time.

Step 4: Configure database adapters

Add adapter configs for your target databases (PostgreSQL, MySQL, SQLite, MongoDB, Redis) in the project config file.

Step 5: Run locally

hyperterse dev

The dev server starts MCP at /mcp and agent routes at /agent/{name}. Use the CLI reference for all available commands.

Deeper Analysis

Architecture choice: Hyperterse sits at the “one process to rule them all” end of the spectrum. If you are currently stitching together a separate agent runtime, an MCP server, a prompt management service, and a data access layer, Hyperterse replaces all four with one compilable unit. The tradeoff is that you are committing to the Hyperterse config model — it is not a polyglot runtime.

When to choose Hyperterse over a custom stack:

• You want A2A agents and MCP tool exposure from the same codebase without maintaining two separate servers
• You need database adapters that come out of the box (PostgreSQL, MySQL, SQLite, MongoDB, Redis)
• You want OpenTelemetry tracing without wiring it up yourself
• Declarative config fits your team’s workflow

Not a good fit if:

• You need a hosted/managed agent runtime (Hyperterse is self-hosted)
• You want to expose non-MCP protocols alongside agents
• You prefer a UI-driven workflow over config files

Practical Evaluation Checklist

• [ ] Run the install script and scaffold a project
• [ ] Define one database-backed MCP tool and one TypeScript handler tool
• [ ] Verify tool resolution and auth execution pipeline
• [ ] Check OpenTelemetry trace output in dev mode
• [ ] Review the docs.hyperterse.com for your specific database adapter

Security Notes

• Per-tool auth (allow_all, api_key, or custom plugin) is configurable but not enforced by default — enable explicitly in your tool configs
• Database adapters use direct connections — apply standard DB access controls (least-privilege DB users, network isolation) in production
• No built-in rate limiting — add at the load balancer or API gateway layer if needed

FAQ

Q: Does Hyperterse replace a full AI backend? A: No. Hyperterse handles agent routing, tool dispatch, and data access — but you still need a model provider (OpenAI, Anthropic, local LLM, etc.) to power the agents themselves.

Q: How does it compare to a raw MCP server? A: An MCP server exposes tools. Hyperterse adds agent orchestration (A2A), prompt templates, resource surfaces, and database adapters — all compiled into one service instead of wired together from multiple packages.

Q: Is there a hosted or managed option? A: Not currently. Hyperterse is self-hosted. Deploy it as a Docker container, a systemd service, or any process manager on your own infrastructure.

Q: What does the production deployment look like? A: Compile your project (hyperterse build), then run the resulting binary behind a reverse proxy. Set HYPERTENSE_ENV=production and configure auth keys for each tool. The binary is standalone — no Node.js runtime needed at deploy time.

Conclusion

Hyperterse solves the “four services to do one job” problem for agentic AI stacks. By combining A2A agent routing, MCP tool exposure, prompt management, and database adapters in one compile-and-deploy process, it cuts operational overhead without sacrificing the protocol surfaces your clients need. It is Apache-2.0, under active development, and worth a look if you are building agentic services that need to touch databases and expose tools to AI clients.

Try it at hyperterse.com, read the full README, or browse the docs for your specific use case.