Surfbots
Cloud-Independent Enterprise AI Knowledge Platform

Turn company knowledge into trusted answers people can use.

Surfbots gives teams a governed AI assistant grounded in approved company knowledge, with source-backed answers, controlled access, conversation continuity, and business-ready outputs. The same Kubernetes-based platform can run locally, on-premises, in a private cloud, or in a public cloud while keeping models, storage, vector search, identity, and infrastructure providers replaceable.

Request a Guided Demo Explore the Platform
Grounded answers + citations Role-aware knowledge access Business-ready outputs Cloud-independent deployment
Find the answerRetrieve the most relevant approved knowledge instead of searching across folders, portals, and disconnected repositories.
Verify the answerUse source references and governed retrieval so important claims can be traced back to approved knowledge.
Control the experienceScope assistants, users, roles, and knowledge boundaries by tenant, business unit, service line, and knowledge base.
Turn answers into workMove useful results into PDF, Word, PowerPoint, Excel, and managed business workflows.

One portable platform, two experiences

Useful for employees. Governable for the teams responsible for AI.

Surfbots combines a focused end-user assistant with an administrative control plane for content, permissions, assistant guidance, reporting, and knowledge lifecycle operations. Both experiences run on the same Kubernetes-based Surfbots Platform so the deployment model is consistent across local, private-cloud, and public-cloud environments.

Surfbots Chat

1
Ask naturallyUsers ask questions in normal language and continue with context-aware follow-ups.
2
Retrieve approved knowledgeRelevant content is retrieved within the user’s authorized business scope.
3
Answer with evidenceThe assistant produces a grounded response with references for verification.
4
Reuse the resultExport useful answers into PDF, Word, PowerPoint, or Excel for downstream work.
Open Chat

Surfbots Admin Studio

Tenant structureOrganize tenants into business units, service lines, and knowledge bases, with scoped administration at each level.
Users + rolesControl who can access which assistant and which knowledge scope.
Knowledge lifecycleExtract documents and OCR locally, choose PII processing, and keep indexes current with incremental source updates and deletions.
Models & SearchOwners can select generation, embedding, and reranking profiles per tenant, see the active models, and tune grounded answers through consistent forms.
Quality + feedbackCapture user feedback, usage signals, and evaluation data for ongoing improvement.
Templates + outputsManage reusable business templates, including fillable PDF workflows.
Responsive administrationGrouped, role-aware navigation collapses to icons on desktop and opens as a drawer on mobile, keeping every available page within reach.
Tenant isolationEach tenant has its own Admin and Chat databases, independent model selections, and dedicated Qdrant collections.
Open Admin Studio

What makes it enterprise-ready

The controls enterprises expect, without losing the simplicity users want.

01

Grounded answers with sources

Responses can include references back to approved files and web content so users can verify important information instead of trusting an opaque answer.

02

Scoped access boundaries

Knowledge visibility can be constrained by tenant, business unit, service line, and assigned role so teams operate inside intentional data boundaries.

03

Portable identity + authorization

Standards-based OAuth2/OIDC identity and role-aware authorization keep access control portable. Keycloak can provide platform-managed identity, while enterprise identity providers can be integrated without coupling the application to one cloud.

04

Conversation continuity

Conversation context can be carried forward so users can ask better follow-up questions without restarting every interaction.

05

Quality and operational signals

Feedback, usage, token metrics, cache signals, and evaluation metadata help teams understand quality, adoption, and opportunities to improve.

06

Business-ready deliverables

Useful responses can move beyond chat through export workflows and centrally managed templates designed for real business processes.

From an answer to an artifact

Surfbots can turn generated responses into formats teams already use for reviews, briefings, analysis, and operational handoffs.

PDF DOCX PPTX XLSX Managed Templates

Designed around business outcomes

A reusable AI layer for the teams that know their domain best.

Each team can have an assistant shaped around its own approved knowledge, terminology, access model, and operating workflows—without forcing the entire company into one undifferentiated chatbot.

Operations

Policy and process answers

Help employees understand procedures, requirements, exceptions, and next steps using controlled operating documentation.

Leadership

Summaries and decision support

Synthesize approved material into executive summaries, comparison views, talking points, and presentation-ready content.

Service Teams

Faster internal support

Give HR, IT, finance, legal, or program teams a consistent first line for answering repeat questions from trusted sources.

Knowledge Owners

Governed reuse at scale

Maintain one controlled knowledge operation while delivering different experiences across teams, functions, and service lines.

Cloud independence, air-gapped deployment, security, and governance

Keep sensitive knowledge entirely inside your environment—with a private AI appliance designed to operate there.

Surfbots Platform is packaged around Kubernetes so the same application architecture can run on a workstation, private data center, isolated network, or managed cloud. For highly sensitive environments, the full RAG pipeline can operate air-gapped with local generation, embedding, and reranking models plus self-hosted vector search, object storage, databases, identity, and ingress. No public cloud AI service is required for the core knowledge workflow when the platform is configured for local providers.

Fully air-gapped AI

Run retrieval and generation without outbound model calls. Local model services handle generation, embeddings, and reranking while knowledge sources, indexes, conversations, and operational data remain inside the isolated environment.

Self-hosted data plane

Qdrant vector search, MinIO-compatible object storage, MongoDB-compatible operational data, PostgreSQL-backed identity services, and Kubernetes networking can run as local platform components instead of external managed services.

Portable identity + authorization

OAuth2/OIDC authentication and Surfbots role/scope resolution support platform-managed identity such as Keycloak or integration with an organization's existing identity provider without coupling authorization to one cloud vendor.

Knowledge boundaries

Retrieval is scoped to approved knowledge associated with the correct organizational domain and role, helping prevent a single global corpus from becoming the default access boundary.

Cloud when you want it

The same containerized application can run on managed Kubernetes in Azure, AWS, Google Cloud, or private infrastructure, while providers for models, storage, identity, and vector search remain replaceable.

Appliance-to-cluster scalability

Start with a turnkey private appliance for a focused deployment, then add capacity or expand into larger Kubernetes infrastructure without replacing the Surfbots application or knowledge model.

Surfbots Private AI Appliance and air-gapped deployment model: When configured with local providers, Surfbots can keep the entire knowledge path inside the controlled network—document ingestion, object storage, vector indexing, retrieval, reranking, prompt construction, model inference, identity, authorization, and generated outputs. Internet connectivity is not required for normal operation after the required container images, model weights, and application artifacts have been staged into the environment.

Private AI without the infrastructure project

A complete Surfbots environment can arrive at your door, connect to your network, and stay inside it.

The Surfbots Private AI Appliance packages the platform into a ready-to-deploy on-premises system for organizations that want the benefits of generative AI without building a GPU stack, exposing sensitive knowledge to external services, or depending on a public-cloud control plane. It is designed for straightforward deployment: receive the appliance, connect power and network, complete organization-specific configuration, load approved knowledge, and begin serving private AI workflows.

01

Delivered as a complete system

Surfbots software, local AI services, vector search, storage, identity, and Kubernetes orchestration can be pre-integrated so customers do not need to assemble the underlying AI infrastructure themselves.

02

Plug into the customer network

Deploy directly inside the organization’s environment and integrate with approved network, identity, and knowledge sources while keeping operational control local.

03

Air-gapped by design

For isolated environments, generation, embeddings, reranking, retrieval, indexes, storage, identity, and application services can all operate locally after release artifacts and model assets are staged.

04

Your knowledge stays with you

Approved documents, vector indexes, conversations, prompts, model inference, and generated outputs can remain within infrastructure controlled by the customer.

05

Start small, scale privately

Begin with a focused departmental or company deployment, then expand capacity by adding private compute nodes or moving the same containerized platform into a larger Kubernetes cluster.

06

One product, multiple deployment models

The same Surfbots application can run as a private appliance, on enterprise Kubernetes, or in a managed cloud—allowing deployment strategy to change without replacing the knowledge platform.

Private AI as an appliance, not a consulting project

Surfbots turns a complex collection of models, retrieval services, databases, identity, storage, and orchestration into a deployable product. The customer focuses on knowledge, governance, and use cases—not assembling an AI infrastructure stack.

Preconfigured On-Premises Air-Gapped Locally Governed

Bring private AI onto your network without building the platform yourself.

Start with a turnkey Surfbots Private AI Appliance for one high-value knowledge workflow, then expand across teams and capacity as adoption grows.

Plan Your Surfbots Demo

Surfbots Dev Platform · New in v0.5.0

Give coding agents current code truth, durable engineering memory, and an independent supervision layer.

Surfbots Dev Platform is a local-first engineering intelligence layer for Cline and other MCP-capable coding agents. It indexes the working repository, preserves the decisions and unresolved work behind it, and adds supervised development checkpoints so substantial changes can be refined, validated, reviewed, and completed against persistent evidence instead of relying only on the coding agent's temporary chat context.

Persistent code intelligence

1
Index the active working treeGit-aware repository discovery, Tree-sitter parsing, code chunking, embeddings, Qdrant vector search, and reranking create a repository-scoped semantic index.
2
Keep context currentManaged snapshots and incremental indexing keep the semantic representation aligned with active development instead of forcing agents to rediscover the codebase from scratch.
3
Expose focused MCP toolsAgents can search code, retrieve files, find symbols and references, inspect repository state, recover development memory, and invoke supervised development workflows.
4
Preserve engineering continuityDevelopment Memory stores repository- and task-scoped decisions, checkpoints, review evidence, validation outcomes, and unresolved work for future sessions.

From scoped task to verified completion

Enforced task lifecycleBegin with a defined repository and file scope, record checkpoints, review the Git changes, verify test evidence, and complete the task.
Aligned Cline + Copilot rulesBundled project rules guide substantial work through the same CLI lifecycle. Start a new agent session after updating the rules.
Context budgets by roleChoose token windows for generation, review, refinement, and collaboration. Local Quality defaults to 32k; recognized OpenAI models default to 128k within known limits.
Memory matched to the modelManaged Qwen defaults to 6 GiB; optional Devstral Small 2 Q4 defaults to 32 GiB. Hosted and external servers manage their own memory.
Project-scoped continuityRepository IDs keep retrieval, memory, and task evidence attached to the right project across sessions. Model profiles remain shared platform settings.
Verified release deliveryv0.5.0 passed 425 automated tests. The bootstrap verifies the release archive checksum, and its manifest identifies the source commit.
Explore the v0.5.0 Updates

Why it matters

Code agents can change. Models can change. Surfbots keeps a persistent engineering layer underneath them: current code evidence, durable project memory, repository-scoped retrieval, and a separate supervisory model that can challenge or validate the implementation before completion.

Code Intelligence Development Memory MCP Configurable AI Supervision Local Models

Request a guided demo

Show us where your teams lose time finding, verifying, or reusing knowledge.

We’ll use that workflow to demonstrate how Surfbots can structure the assistant, knowledge scope, governance model, and business outputs around a real operating need.

Good starting points

  • Employees repeatedly searching across policies, procedures, or internal documentation.
  • Teams spending time summarizing the same information for different audiences.
  • Business units that need AI assistance but cannot share all knowledge with every user.
  • Organizations that want centralized AI governance without creating one generic enterprise assistant.
  • Knowledge owners who want feedback and quality signals tied to real assistant usage.