Skip to content

The platform

trusted answers and AI agents.

Turn your company's data into accurate answers and AI agents that get work done, with guardrails and logs built in. Start with our SaaS today, or run it on your own infrastructure.

Start nowTalk to an engineer

How it works

Guardrails, model choice and deployment options come standard. Turn them on and tune them to your needs.

  • Guardrails on every call

    Built-in guardrails check every question and every answer. The same rules apply to a chat window, an API call or a scheduled agent.

  • Use the models you choose

    Add one model or several, and NeuralSeek balances the load across them. Pick which jobs each model does. Your data, guardrails and apps stay where they are.

  • Runs where you need it

    Use our SaaS, or run it in your own cloud account, on your OpenShift cluster, or air-gapped on your own hardware. The guardrails and features are the same everywhere.

What's inside

Everything you need to build with AI

Each part works with the others. A guardrail you set for answers also applies when an agent uses them, whichever model answers and wherever it runs.

  • Seek: answers from your data

    Answers drawn from your own documents, scored for confidence and coverage, with the source passages highlighted. Advanced search and a cache of approved answers are built in.

    Explore Seek
    console › seek
    An answer in the NeuralSeek console with its sourced passages highlighted, beside the session panel and a table of match, confidence and coverage readings.

  • mAIstro: AI agents

    Build multi-step agents in a visual editor, or in code with NTL, NeuralSeek's template language. Pick a model for each step, branch on results, run steps in parallel and schedule agents to run on their own.

    Explore mAIstro
    console › maistro › agent editor
    The mAIstro agent editor: a node graph running from a message search through PII removal and a protection check to an LLM plan-and-act pair and a presentation output, with the function library alongside.

  • Guardrails

    Hundreds of guardrails screen questions on the way in and check answers on the way out. Set them per topic, and run red-team tests on your own agents anytime.

    Explore guardrails
    console › neural config › guardrails
    The Guardrails panel of Neural Config, its tabs running from semantic scoring, prompt injection, PII and profanity to custom governance, open on Min Confidence: a threshold slider, a custom-message agent and a fallback reply.

  • Governance

    A record of everything that ran: answer scores over time, agent runs with their timings, and a change log for your settings that lets you roll back.

    Explore governance
    console › governance › overview
    The Governance overview: tiles for semantic confidence, question resolution, the hate-abuse-profanity block rate, prompt-injection risk and questions containing PII, over a chosen date range.

  • Connectors

    Ready-made nodes for your databases, file stores, search and document libraries. Each one is a step in an agent, and every agent can be called over REST or as an MCP tool.

    Explore connectors
    console › maistro › function library
    Connector cards from the mAIstro function library laid out in a grid: database, search, file-storage and document-library nodes, each with a one-line description and a settings control.

  • Bring any model

    Add an LLM by choosing a platform and a model, and see up front which functions it supports. Add several and NeuralSeek balances the load. Swapping a model is a settings change.

    See model options
    console › neural config › add an llm
    The Add an LLM dialog: OpenAI chosen as the platform and GPT-4o as the model, beside the model's notes, including the one function it does not support.

  • Deployment

    One container with everything above. Start with our SaaS, buy through the AWS, Azure or IBM Cloud marketplaces, run it on IBM Fusion, or install it on your own OpenShift or Kubernetes, fully isolated if you need it.

    Explore deployment

Common questions

What teams usually ask before they start.

  1. How is this different from a basic RAG setup?

    Retrieval is one part. NeuralSeek adds checks on every call, a full record of each answer you can look up months later, agents that keep working when you change models, and deployment inside your own environment.

  2. Does this lock us to one model vendor?

    No. You choose which jobs each model does, pick a model for each agent step, and can override the model on a single API call. Changing any of them is a settings change.

  3. We already have a model provider and a vector store. Where does this fit?

    On top of both. Use your own provider keys and point NeuralSeek at the search you already run. It adds grounding checks, guardrails, agents and a full audit trail.

Start building with NeuralSeek today.

Sign up online and start with a two-week trial, or talk to a NeuralSeek expert about running it in your own cloud.

Start nowRequest a demo