v2.5.0 is out. With it, Hyperterse is now - the agentic server framework!
Building up from the previous release, Hyperterse now implements the full A2A spec along with the full MCP spec.
This means agents, tools, resources, and prompts, are all first-class citizens!
v2.5.0 is out. With it, Hyperterse is now - the agentic server framework!
Building up from the previous release, Hyperterse now implements the full A2A spec along with the full MCP spec.
This means agents, tools, resources, and prompts, are all first-class citizens!
I am very please to release something I've been working on.
One framework. Fully A2A and MCP specs compliant.
Same declarative configs, same powerful engine.
Excited for you to try it out!
this is actually an interesting build.
had a lot of fun building this.
one especially noteworthy thing is - this gives you an a2a (agent to agent) protocol compatible endpoint.
which means hyperterse agents can talk to each other and to other agents.
with a simple config.
🚀 v2.4.0: declarative agents in the engine - discovery, model providers (OpenAI / Gemini / Vertex / OpenAI-compatible), tool access policies, and full docs.
Ship agents where you already ship MCP tools.
Upgrade: https://t.co/AHtZv96YVP
Full changelog: https://t.co/fS0eNdACHq
We are live on Product Hunt! 🚀
Hyperterse v2 is a ground-up rebuild to be the most efficient MCP framework out there.
Go ahead and upvote - would love your support!
👉 https://t.co/AcGidOtyWD
Be me
> keep trying to provide ability to agents to query my data
> try text-2-sql, only rag, rl over foundational models
> get frustrated, discover mcp
> build simple tool to connect databases to agents over mcp
> realise it is good but not great
> launch it anyway and get some initial traction
> hit same problems figure out this needs a rewrite
> create a framework for mcps
> database connectors, custom handlers, js runtime, dynamic tool discovery all baked in
> mcps infinitely scalable - problem solved
> finally launch on ph, 3 months after new version released
Go support it
I am so excited to launch @hyperterse on Product Hunt tomorrow. I think it is the framework for anyone wanting to create MCPs. It is extremely powerful and "just works".
@garrytan You should really think about giving @hyperterse a try to build your MCPs. It is extremely fast + completely config-based, allowing you to dip into custom handlers whenever required. It still only gives you search and execute tools to your LLMs so no token bloat.
Swapping any provider anytime is a huge win, first step of many to something we want to eventually establish as a protocol!
Sandboxes are super powerful, and accessing them should be uniform.
Spent a lot of time tightening the core loop in this one.
The goal is simple:
You shouldn’t be writing MCP servers.
You should be declaring intent.
v2.3.0 gets us meaningfully closer.
- SQLite connector now natively supported
- Ability to create prompts, resources and templates
⚡️ v2.3.0 just dropped.
The goal is simple:
You shouldn’t build MCP servers anymore. You should just describe them.
Everything else? Handled
- Full adherence to MCP spec
- SQLite connector
Upgrade: https://t.co/AHtZv96YVP
Full changelog: https://t.co/Nztu5SBL3w