Omni Nli
Run natural language inference (entailment, contradiction, neutral judgments) from your coding agent via a local or self-hosted MCP HTTP endpoint.
Overview
Omni NLI MCP is a Build-phase MCP server that provides natural language inference over HTTP so agents can obtain structured semantic judgments between text pairs.
What is this MCP server?
- MCP server focused on natural language inference workflows (package omni-nli on PyPI, v0.1.1).
- Uses streamable-http transport at http://127.0.0.1:8000/mcp/ for local integration.
- Lets agents outsource pairwise or premise-hypothesis style semantic checks instead of prompting raw LLM guesses.
- Fits builder-side eval scripts, moderation prototypes, and RAG guardrails that need explicit NLI labels.
- Repository hosted at CogitatorTech/omni-nli on GitHub for self-host customization.
- MCP server version 0.1.1
- Default endpoint http://127.0.0.1:8000/mcp/
- Transport type streamable-http via PyPI package omni-nli
Community signal: 3 GitHub stars.
What problem does it solve?
Builders hand-roll brittle prompt hacks for entailment checks instead of a dedicated NLI endpoint their agent and tests can share.
Who is it for?
Developers adding semantic consistency checks, lightweight eval harnesses, or agent guardrails with a self-hosted NLI MCP service.
Skip if: Non-technical users who want hosted NLI without running a local server, or teams that only need generic summarization without inference labels.
What do I get? / Deliverables
Your agent calls a streamable-http NLI MCP service for consistent inference labels you can plug into evals, RAG filters, or workflow gates.
- HTTP MCP endpoint for NLI requests from coding agents
- Structured semantic labels usable in automated tests or agent branches
- Local dev loop for NLI without bespoke REST wrappers
Recommended MCP Servers
Journey fit
Canonical shelf is Build because NLI is typically embedded in product logic, eval harnesses, or agent pipelines you implement—not in early competitor spreadsheets. Agent-tooling captures an MCP microservice that agents call for structured semantic judgments during feature or eval work.
How it compares
Dedicated NLI MCP microservice, not a general-purpose chat completion plugin.
Common Questions / FAQ
Who is omni-nli for?
AI-forward solo builders and small teams who run local MCP services and need natural language inference in agent or product pipelines.
When should I use omni-nli?
Use it while building features or eval tooling that require entailment, contradiction, or neutral classifications between utterances.
How do I add omni-nli to my agent?
Start the omni-nli PyPI service, register the streamable-http remote URL http://127.0.0.1:8000/mcp/ in your MCP client, and ensure port 8000 is reachable from the agent runtime.