Skip to content

📦 Installation & Setup

Choose the right installation path for your workflow: zero-install agent runner, one-line standalone binary, package manager, container, or client SDK.

1. Zero-Install MCP Runner (for AI Agents)

If you are using Claude Desktop, Cursor, Windsurf, or Claude Code:

npx -y @spectrayan/spector mcp

The NPX launcher connects to a running local Spector Synapse daemon on :7070 if healthy. If no daemon is running, it downloads spector.jar to run an in-process memory kernel with embedded ONNX vector embeddings (requires OpenJDK 25+).

2. One-Line Standalone Installers

Install the standalone spector CLI binary directly to ~/.spector/bin and update your system PATH:

curl -fsSL https://raw.githubusercontent.com/spectrayan/spector/main/scripts/install.sh | sh

irm https://raw.githubusercontent.com/spectrayan/spector/main/scripts/install.ps1 | iex

Verify the installation: 

spector doctor

3. Package Managers (Available with Release Tags)

Note

Homebrew and Scoop package manager installations pull spector.jar from GitHub Releases. They activate once the release binary is published for your tag.

# Tap the official repository and install
brew tap spectrayan/spector https://github.com/spectrayan/spector
brew install spector

scoop install https://raw.githubusercontent.com/spectrayan/spector/main/packaging/scoop/spector.json

4. Docker & Docker Compose

Launch the complete cognitive stack (core memory engine on :7070 and Cortex 3D neural explorer on :7700) with zero local prerequisites:

# Clone the repository
git clone https://github.com/spectrayan/spector.git
cd spector

# Launch engine and Cortex dashboard
docker compose up -d

See the Docker Deployment Guide for multi-stage build instructions and GPU acceleration.

5. Client SDKs (Zero Java Required)

If you are building applications that interact with Spector over HTTP/SSE:

pip install spector-client
See the Python SDK Guide.

npm install @spectrayan/spector-client
See the TypeScript SDK Guide.

<dependency>
    <groupId>com.spectrayan</groupId>
    <artifactId>spector-client</artifactId>
    <version>0.1.0-beta</version>
</dependency>

Note

Java Embed Dependency: Applications should depend on com.spectrayan:spector-client. The root coordinate com.spectrayan:spector is the multi-module reactor parent POM and does not contain client classes.

See the Java Client Guide.

6. Building from Source (Engine Contributors)

System Requirements

Requirement Minimum Recommended
JDK OpenJDK 25+ (with Vector API & FFM) Eclipse Temurin 25
Maven 3.9+ 3.9.9
RAM 1 GB 4 GB+
Architecture x86_64 (AVX2/AVX-512) or aarch64 (NEON) SIMD hardware support

Verified architectures

Both listed architectures are built and tested on every commit:

Architecture CI runner Covers
x86_64 ubuntu-latest Full reactor, plus reproducibility, dependency pinning, coverage and provenance checks
aarch64 ubuntu-24.04-arm Full reactor and test suite

This matters more than a support matrix usually does. Spector's hot paths use the Java Vector API and off-heap MemorySegment layouts, where vector lane counts, alignment and byte order differ between architectures. Until #992, CI ran x86_64 only — and the INT8 quantisation kernels were entirely unusable on aarch64 while this page already claimed aarch64 support. Both architectures are now tested, so the claim on the row above is checked rather than asserted.

Other aarch64 platforms — Apple Silicon and AWS Graviton in particular — run the same NEON path as the CI runner, but are not themselves part of the matrix.

Build Instructions

# 1. Clone the repository
git clone https://github.com/spectrayan/spector.git
cd spector

# 2. Build the reactor and package spector.jar
mvn clean package -pl synapse/spector-cli -am -DskipTests

# 3. Launch the standalone CLI
java --add-modules jdk.incubator.vector \
     --enable-native-access=ALL-UNNAMED --enable-preview \
     -jar synapse/spector-cli/target/spector.jar doctor