About
Kimberlite MCP server for LLM and third-party integrations
README
Kimberlite
Crates.io Downloads Documentation Rust Edition License CI VOPR Fuzz Nightly Formal Verification Discord
A verifiable database for healthcare.
Built for clinical and digital-health systems where PHI integrity is non-negotiable.
🔬 Developer Preview - Explore deterministic database concepts through production-quality code
Kimberlite is a verifiable, durable database engine designed for environments where data integrity, auditability, and trust are non-negotiable. Built around a single principle:
All data is an immutable, ordered log. All state is a derived view.
Why Kimberlite?
The compliance tax is real. In healthcare, you're forced to build:
- Immutable audit trails for every change
- Cryptographic proof of data integrity
- Per-tenant encryption and isolation
- Point-in-time reconstruction
Most teams bolt these onto existing databases. Kimberlite builds them in.
Key approach:
- Immutable audit trail - Hash-chained append-only log means every action is recorded
- Time-travel queries - Reconstruct any point-in-time state via MVCC (
AT OFFSET nandAS OF TIMESTAMP '...'both shipped) - Multi-tenant isolation - Cryptographic boundaries prevent cross-tenant access
- Multi-layer verification - TLA+ protocol specs, Coq crypto proofs, Alloy structural models, Ivy Byzantine invariants, Kani bounded model checking, MIRI UB detection (details)
Designed for healthcare: EHR-adjacent systems, digital-health platforms, payer/RCM workloads, and clinical research — every primitive is sized to HIPAA, with GDPR (international health) and SOC 2 / FedRAMP (clinical SaaS / federal health) as load-bearing extensions.
Who Should Explore This
- 🏥 Clinical IT teams - Build HIPAA-ready EHR-adjacent systems with built-in audit trails
- 🚀 Digital-health founders - Ship FHIR-aware apps without bolting on the compliance plumbing
- 💳 Payer & RCM engineers - Append-only claims ledgers with cryptographic chain-of-custody
- 🔬 Clinical researchers & informaticists - 21 CFR Part 11-friendly audit, de-identification, and time-travel queries
Perfect for learning. Not yet recommended for production deployments (see Status below).
Quick Start
5-minute quickstart: See Getting Started for a complete tutorial with explanations.
TL;DR:
# Install (see docs/start/installation.md for all options)
curl -fsSL https://kimberlite.dev/install.sh | sh
# Initialize (or: kimberlite init for interactive wizard)
kimberlite init myproject
kimberlite dev
# Studio: http://localhost:5555, DB: 127.0.0.1:5432
Try time-travel queries:
CREATE TABLE patients (id INTEGER, name TEXT);
INSERT INTO patients VALUES (1, 'Alice'), (2, 'Bob');
-- View current state
SELECT * FROM patients;
-- View state as of a specific log offset (MVCC time-travel)
SELECT * FROM patients AT OFFSET 0;
-- Or as of a wall-clock timestamp (resolved via the audit-log index)
SELECT * FROM patients AS OF TIMESTAMP '2026-01-15T00:00:00Z';
Documentation
- Quick Start - Get running in 90 seconds
- CLI Reference - All commands
- SQL Reference - Supported SQL syntax
- Architecture - How Kimberlite works
- Roadmap - Future features and enhancements
- Changelog - Release history and completed work
- Contributing - How to contribute
Building from Source
# Clone and build
git clone https://github.com/kimberlitedb/kimberlite.git
cd kimberlite
cargo build --release -p kimberlite-cli
# Binary is at ./target/release/kimberlite
Development Commands
just build # Debug build
just build-release # Release build
just test # Run all tests
just nextest # Faster test runner
just clippy # Linting
just pre-commit # Run before committing
Key Features
What Makes Kimberlite Unique:
- ✅ Immutable audit trail - Hash-chained append-only log (SHA-256 for compliance, BLAKE3 for performance)
- ✅ Time-travel queries - MVCC enables both
AT OFFSET nandAS OF TIMESTAMP '...' - ✅ Deterministic core - Functional Core / Imperative Shell pattern enables perfect replication
- ✅ Multi-tenant isolation - Per-tenant storage with cryptographic boundaries
- ✅ Multi-layer verification - TLA+ protocol specs (TLC in PR CI, TLAPS nightly), Coq crypto proofs, Alloy structural models, Ivy Byzantine invariants, Kani bounded model checking, MIRI undefined-behavior detection (details)
- ✅ SQL interface - SELECT with aggregates, GROUP BY/HAVING, UNION/INTERSECT/EXCEPT, all JOIN kinds, CTEs (incl. recursive), subqueries (incl. correlated), window functions, JSON operators, scalar expressions.
INSERT ... ON CONFLICTupsert.ALTER TABLE ADD/DROP COLUMN. Multi-statement transactions planned post-v1.0. - ✅ Compliance surface -
client.compliance.{eraseSubject, audit.query, export_subject, breach_*, consent.{grant,revoke}}reachable from all SDKs with signed witnesses, hash-chain tamper-evidence, and GDPR Art 6 consent basis - ✅ Column-level masking policies -
CREATE MASKING POLICYDDL composes with RBAC + break-glass - ✅ Tamper-evidence - CRC32 checksums + hash chains detect corruption
- ✅ Viewstamped Replication (VSR) - Full multi-node consensus (Normal, ViewChange, Recovery, Repair, StateTransfer, Reconfiguration)
- ✅ RBAC/ABAC enforcement - Per-role row/column filters; HIPAA TPO + break-glass policies built-in
- ✅ Security hardened - pre-launch audit completed; message signatures, replay protection, DoS limits, zero unallowed RustSec advisories
See CHANGELOG.md for per-release detail; ROADMAP.md for what's next.
Use Cases
Kimberlite is built for healthcare:
- EHR / EMR systems — patient records, clinical notes, encounters with FHIR R4 and HL7 v2 native types
- Payer / RCM — X12 837 claim ingestion, 835 remittance reconciliation, audit-by-default
- Clinical research — IRB-aware retention, consent-bound queries, Safe Harbor de-identification for secondary use
- Digital health — multi-tenant SaaS with per-tenant encryption and BAA-ready audit chain
Examples
See the examples/ directory for:
- quickstart/ - Getting started
- rust/ - Rust SDK examples
- docker/ - Docker deployments
- healthcare/ - HIPAA-ready schema
Architecture
┌──────────────────────────────────────────────────┐
│ Kernel (pure state machine: Cmd -> State + FX) │
├──────────────────────────────────────────────────┤
│ Append-Only Log (hash-chained, CRC32) │
├──────────────────────────────────────────────────┤
│ Crypto (SHA-256, BLAKE3, AES-256-GCM, Ed25519) │
└──────────────────────────────────────────────────┘
See docs/concepts/architecture.md for details.
Why Kimberlite vs. Traditional Databases?
| Feature | PostgreSQL | Kimberlite |
|---|---|---|
| Data model | Mutable tables | Immutable log + derived views |
| Audit trail | Manual triggers | Built-in (every write logged) |
| Time-travel | Extensions (complex) | Native SQL (AS OF TIMESTAMP) |
| Integrity | Checksums | Hash chains + CRC32 |
| Consensus | Streaming replication | VSR (deterministic, multi-node) |
| Best for | General OLTP | Compliance-heavy workloads |
Trade-offs: Kimberlite trades some write throughput for built-in auditability and tamper-evidence. Quantitative re-baseline against current hardware is a v0.9.0 target; see FAQ for the qualitative comparison.
Learning Resources
Documentation Deep Dive
- docs/concepts/architecture.md - FCIS pattern, kernel design, consensus
- docs/internals/testing/assertions-inventory.md - Production assertion policy + paired
#[should_panic]tests - docs/internals/testing/overview.md - VOPR deterministic simulation testing
- docs/concepts/pressurecraft.md - Code quality standards
- docs/concepts/compliance.md - HIPAA-ready, SOC 2-ready, GDPR-ready patterns
Community
- 💬 Discord - Join for real-time support, design discussions, and community
- 📖 Documentation - Comprehensive architecture and usage guides
- 🐛 Issues - Bug reports and feature requests
- 💡 Discussions - Questions, ideas, and design conversations
- ❓ FAQ - Frequently asked questions
Status
v0.x — Developer Preview. Stable enough for prototypes, learning, internal tools, and compliance research. Not yet battle-tested at scale.
- ✅ Core is solid: 3,000+ tests, deterministic simulation, production-grade crypto.
- ✅ Architecture is stable: FCIS pattern, immutable log, full multi-node VSR consensus.
- ✅ Security hardened: pre-launch audit closed, zero unallowed RustSec advisories, supply-chain pins.
- ✅ SDKs are production-grade: Rust, TypeScript, and Python SDKs ship full data-plane + compliance + admin surface, with connection pooling and real-time subscriptions. See SDK parity matrix.
- ⚠️ Wire protocol may still evolve between minor versions. See CHANGELOG.md for the current version and any breaking changes.
Use for: internal tools, prototypes, learning database internals, compliance research.
Wait for v1.0 if you need: API stability guarantees, large-scale production deployment, commercial support, or third-party SOC 2 / HIPAA / GDPR attestations. v1.0 is checklist-gated with no fixed date — see ROADMAP.md for the gates.
Post-v1.0: a managed cloud service (Kimberlite Cloud) is planned alongside the OSS core. The core stays OSS; the cloud adds ops, scaling, and compliance-ready shared-responsibility — similar to CockroachDB Serverless on top of CockroachDB OSS.
Known Limitations (v0.8.0)
We surface these up front so you don't trip over them:
- Single-node only for production. Multi-node VSR consensus is implemented end-to-end and PR-gated against TLA+ specs, but the
kimberlite-clustercrate is marked "not ready for public use" — production multi-node is a v1.0 gate. Single-node is stable. - No multi-statement transactions.
BEGIN/COMMIT/ROLLBACKare planned post-v1.0; the parser rejects them today. Single statements are atomic. For cross-table workflows, see the outbox-pattern recipe in examples/cookbook/. NOW()/CURRENT_TIMESTAMP/CURRENT_DATEpanic if reached unfolded. v0.7.0 shipped the sentinel variants; the planner-sidefold_time_constantspass is a v0.9.0 in-flight item. Until then, pass an explicit timestamp parameter from the application instead.- No published performance baselines yet. Hardware re-baseline is a v0.9.0 target. Correctness is covered by 3,000+ tests + VOPR deterministic simulation; throughput / latency numbers are coming.
- Wire protocol may evolve between minor versions. Breaking changes are called out in CHANGELOG.md. Pin server and SDK to compatible versions until v1.0.
- Go SDK Phase 1 lands in v0.9.0. Rust / TypeScript / Python are at full data-plane + compliance-surface parity today.
See ROADMAP.md for the full list of in-flight and deferred items, and the v1.0 readiness gates.
Compliance posture
Kimberlite ships the cryptographic and architectural primitives that HIPAA demands, with GDPR / SOC 2 / ISO 27001 / FedRAMP as overlays: hash-chained immutable audit log (server-walked attestation via audit.verifyChain()), AES-256-GCM at rest, Ed25519 signed exports, multi-tenant key isolation, GDPR Article 17 erasure orchestration with requestId correlation, GDPR Article 6 consent-basis tracking on the wire, column-level masking policies that compose with RBAC + break-glass, Safe Harbor de-identification of all 18 §164.514(b)(2) identifiers, and healthcare-relevant frameworks formally modelled (TLA+ + Coq + Kani + VOPR scenarios) — see docs/compliance/certification-package.md.
What we have NOT done: third-party SOC 2 Type II audit, HIPAA attestation + BAA, FedRAMP authorization, GDPR readiness review. These are v1.0 gates. Until then, Kimberlite enables your compliance posture; certification remains your team's responsibility (BAA with your hosting provider, pen test, audit firm engagement, internal controls). If you're a regulated startup, you can build on Kimberlite today knowing the substrate is sound — but the audit letters and certifications are work you (or we, post-v1.0) still have to do.
SDKs
Kimberlite provides idiomatic client libraries for multiple languages:
| Language | Status | Package | Install |
|---|---|---|---|
| Rust | ✅ Ready | kimberlite-client |
cargo add kimberlite-client |
| TypeScript | ✅ Ready (Node 18/20/22/24, prebuilt napi) | @kimberlitedb/client |
npm install @kimberlitedb/client |
| Python | ✅ Ready | kimberlite |
pip install kimberlite |
| Go | 📋 Planned (v0.9.0 Phase 1) | — | See ROADMAP |
| Java | 📋 Planned (v1.0 gate) | com.kimberlite:kimberlite-client |
Maven / Gradle |
| C++ | 📋 Planned (v1.0 gate, via FFI) | kimberlite-cpp |
Coming soon |
See docs/reference/sdk/overview.md for architecture and docs/reference/protocol.md for wire protocol specification.
License
Apache 2.0
Contributing
- Read CLAUDE.md for development guidelines
- Review docs/concepts/pressurecraft.md for coding standards
- Open issues for design discussions
Installing Kimberlite
This server has no published package — it is built from source. Open the repository and follow its README.
▸ github.com/kimberlitedb/kimberliteFAQ
Is Kimberlite MCP free?
Yes, Kimberlite MCP is free — one-click install via Unyly at no cost.
Does Kimberlite need an API key?
No, Kimberlite runs without API keys or environment variables.
Is Kimberlite hosted or self-hosted?
Self-hosted: the server runs locally on your machine via the install command above.
How do I install Kimberlite in Claude Desktop, Claude Code or Cursor?
Open Kimberlite on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
Fetch
Web content fetching and conversion for efficient LLM usage.
AWS KB Retrieval
Retrieval from AWS Knowledge Base using Bedrock Agent Runtime.
by modelcontextprotocolSpring AI MCP Server
Provides auto-configuration for setting up an MCP server in Spring Boot applications.
llm-analysis-assistant
A very streamlined mcp client that supports calling and monitoring stdio/sse/streamableHttp, and can also view request responses through the /logs page. It also
by xuzexin-hzMCP-Agent
A simple, composable framework to build agents using Model Context Protocol by [LastMile AI](https://www.lastmileai.dev)
by lastmile-aiSpring AI MCP Client
Provides auto-configuration for MCP client functionality in Spring Boot applications.
mcp.natoma.ai
A Hosted MCP Platform to discover, install, manage and deploy MCP servers by [Natoma Labs](https://www.natoma.ai)
MCPHub
Website to list high quality MCP servers and reviews by real users. Also provide online chatbot for popular LLM models with MCP server support.
MCP Servers Rating and User Reviews
Website to rate MCP servers, write authentic user reviews, and [search engine for agent & mcp](http://www.deepnlp.org/search/agent)
mkinf
An Open Source registry of hosted MCP Servers to accelerate AI agent workflows.
Compare Kimberlite with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All ai MCPs
