Encapsa

AI-Native Vertical SaaS. Privacy-Native by Design. Enterprise grade AI from the Platform Up.

Enterprise AI that ships as applied product — not raw model access. This organization is where we open source the developer tools, agents, and libraries pulled out of the Encapsa AI platform.

Encapsa is an AI-native vertical SaaS platform with 300+ live enterprise deployments across seven industries. We replace what most enterprises otherwise buy from six-plus AI vendors — foundation-model access, a horizontal assistant, NLP-to-SQL reporting, deep research, document AI, connectors, workflow automation, and the full digital-presence layer — with a single, unified platform that runs on a closed-circuit, privacy-native architecture. Client data never has to leave their environment.

This GitHub organization is the developer face of that platform. Everything we publish here has already been battle-tested inside Encapsa. If a tool has earned its place in our stack, we open source it here so the rest of the developer community can use it too.

Our high-performance website and UI libraries live in a separate org: github.com/opensite-ai. This org (encapsa-ai) is where we publish everything else — developer tooling, AI utilities, agent components, connectors, CLIs, and desktop apps.

The libraries and tools in this org back real product surfaces. Here's a look at what they power.

Ingest from databases, SOPs, financials, websites, and flat files. Generate reports, decks, audio overviews, onboarding kits, and shareable AI agents — every output grounded in your connected data and cited inline.

Ask business questions in plain English. Get charts, tables, and narrative summaries back in seconds. Full CRUD via NLP, cross-location analysis, and scheduled reporting — driven by a proprietary multi-model orchestration layer.

Track and grow visibility inside ChatGPT, Perplexity, Google SGE, and Gemini. Real-time citation share, tracked prompts, and structured-data automation for the answer-engine era.

Purpose-built agents orchestrated across connected systems. This is insurance underwriting: 16 steps, from inbound submission through ACORD parsing, entity verification, KYB screening, and delivered underwriting memo — all coordinated across connected AMS, CRM, and third-party data providers.

Natural-language deck generation, data-connected charts, template-based refresh, and Deep Research integration for external benchmarks. This is a live enrollment-analytics board deck pulled from a real client database.

Automated reputation defense across 20+ platforms. Brand-voice AI replies, challenge scoring, sentiment trend analysis, and multi-location dashboards — orchestrated by a supervising reputation agent.

Traffic trends, source attribution, and AI-answer-engine share, exposed through a conversational analytics agent so operators can ask questions instead of building dashboards.

┌───────────────────────────────────────────────────────────────────┐
│  AI Agent Orchestration                                           │
│  Project Manager · Designer · Content · Trend · Reputation        │
├───────────────────────────────────────────────────────────────────┤
│  Multi-Model Router                                               │
│  Dynamic routing per task · reporting · research · generation     │
├───────────────────────────────────────────────────────────────────┤
│  Reasoning Retrieval                                              │
│  RAG · PageIndex · proprietary vector database                    │
├───────────────────────────────────────────────────────────────────┤
│  Connector Engine                                                 │
│  ERP · CRM · POS · SQL · SFTP · Industry APIs (bidirectional)     │
├───────────────────────────────────────────────────────────────────┤
│  Closed-Circuit LLM                                               │
│  Shared · Single-Tenant · On-Premise / Private Cloud              │
└───────────────────────────────────────────────────────────────────┘

Everything above the connector layer is where our open source work lives — orchestration primitives, retrieval utilities, agent components, and developer tooling. The layers below stay proprietary because that's where our clients' data lives.

Every module we open source has been proven inside at least one of these verticals in production:

This org is intentionally scoped to developer-facing work — CLIs, desktop apps, agent primitives, retrieval utilities, connector SDKs, and MCP servers.

• dep-sync — Local-first desktop app for visualizing and resolving version drift across internal npm packages. No registry lookups, no network calls. Tauri + Rust + React 19. Reads your package.json files, computes a dependency-first update order via Kahn's algorithm, and previews the exact package.json diff before you apply it. Exports an AI-agent context file for downstream coding agents.

  

• Connector SDKs — TypeScript and Rust bindings for the Encapsa Connector Engine.
• Agent primitives — Building blocks pulled from our 50+ specialized agent implementations.
• Retrieval utilities — Reasoning-based retrieval helpers, PageIndex components, and evaluation harnesses.
• MCP servers — Model Context Protocol servers for Encapsa data sources and workflow triggers.
• CLIs and desktop apps — More local-first developer tools in the shape of dep-sync.

Watch the org (top-right on this page) if you want to see them the moment they ship.

Looking for our web performance libraries, React hooks, Semantic UI Engine, form library, and Ruby/Rust utilities? Those live in @opensite-ai, powering the Octane™ website platform.

Every project we release here has to hold up against the same set of rules we apply internally:

• Local-first when possible. No telemetry, no network calls, no phone-home — unless a network round-trip is the whole point of the tool.
• Boring, correct primitives. Rust's semver, Kahn's algorithm, standard graph structures. We prefer battle-tested crates and packages over clever homegrown reinventions.
• AI-aware, not AI-hyped. Our tools should make working with AI agents better — clean context exports, deterministic outputs, structured JSON, agent-friendly diffs — without pretending that everything needs an LLM inside it.
• Diff-clean. Anything that writes files preserves key order, indentation, trailing newlines, and range style. If it touches your repo, it has to earn a clean review.
• Cross-platform by default. macOS (Apple Silicon and Intel), Linux, and Windows are all first-class targets.
• Permissive licensing. MIT unless there's a specific reason to pick otherwise.

If you found this org through the platform side of Encapsa — you're evaluating a private-labeled AI platform, an on-premise deployment, or a vertical build-out — here's the shortest path:

• Read the platform overview for the product tour.
• Review security and deployment options — shared multi-tenant, dedicated single-tenant, and fully on-premise with local inference.
• Reach out for a scoped conversation about your vertical, data residency requirements, and integration surface.

Data posture, briefly:

• ✅ Closed-circuit LLM — client data isolated, never used to train shared models.
• ✅ Dedicated single-tenant option for regulated workloads (healthcare, insurance, financial services).
• ✅ Fully on-premise option — Encapsa deployed inside the client's security perimeter with local inference.
• ✅ Contractual commitments in the MSA/DPA, not just marketing copy.

We welcome contributions on every repository in this org. General principles:

1. Fork the repository you want to contribute to.
2. Create a feature branch (git checkout -b feature/your-feature).
3. Commit your changes with a clear, imperative message.
4. Push and open a pull request against main (or the repository's stated default branch).
5. Explain the why — link an issue or describe the problem you're solving. Working code is easier to accept when the intent is clear.

Individual repositories may add their own CONTRIBUTING.md, coding conventions, or CLA. Follow the per-repo rules first.

• Bugs: include reproduction steps, expected vs. actual behavior, and environment details (OS, language version, package version).
• Features: describe the use case first, the proposed API second.
• Security: please do not file public issues for security concerns. Email us via the contact on encapsa.ai so we can triage responsibly.

• Platform: encapsa.ai
• Hospitality vertical: dashtrack.com
• Sister org (web performance libraries): github.com/opensite-ai
• Hugging Face: huggingface.co/encapsa
• LinkedIn: linkedin.com/company/encapsa

Unless a repository states otherwise, everything published in this organization is released under the MIT License. Check the LICENSE file in each repo for specifics.