AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Home/
LLM Infra/
osaurus
osaurus logo

osaurus

Active·★ 8.1k·MIT·Updated 2026-10-27

Osaurus, created by Dinoki Labs, calls itself the AI edge runtime for macOS: a native, always-on foundation for AI apps and workflows on Apple Silicon. It runs local models with MLX and connects to remote providers (Anthropic, OpenAI, OpenRouter, Ollama, LM Studio or any compatible API), and it exposes OpenAI-, Anthropic- and Ollama-compatible endpoints so existing tools can use it as a drop-in server.

Beyond inference, the README lists an MCP server that exposes tools to agents, remote MCP providers whose tools can be aggregated, a plugin system with a community registry, custom agents with their own prompts, tools and themes, a cryptographic identity system for humans and agents, a four-layer memory system (profile, working memory, summaries and a knowledge graph), skills compatible with the Agent Skills format that can be imported from GitHub or files, and schedules for recurring AI tasks. Install with `brew install --cask osaurus` or from the releases page, then launch from Spotlight or run `osaurus ui`. It needs macOS 15.5 or later (Apple Foundation Models need macOS 26) and is MIT licensed.

This write-up is based on the project's README and public repository data. We have not hand-tested the tool. Data as of 2026-10-24.

License
MIT
Latest release
0.25.26 · 2026-10-24
Last commit
2026-10-24
Contributors
41
Forks
483
Open issues and PRs
220
First published
2025-08-17

Source: GitHub, npm and PyPI public APIs.

#AI Edge Runtime#macOS Native#Local LLM Inference#API Gateway#Agent Framework#Plugin System#On-device AI#Speech-to-text
$ Install
$ brew install --cask osaurus
↗ Visit site★ GitHub
01

Features

01Local LLM Inference with MLX Runtime for Apple Silicon
02Multi-API Compatibility (OpenAI, Anthropic, Ollama, etc.)
03MCP Server for AI Agent Tooling and Remote MCP Providers aggregation
04Advanced 4-layer persistent memory system with hybrid search
05Integrated on-device Voice Input and Transcription with VAD Mode
02

Why choose it

+Runs local models with MLX on Apple Silicon and also connects to cloud providers from the same app.
+Drop-in OpenAI, Anthropic and Ollama compatible endpoints let existing tools use it without changes.
+Built-in MCP server plus aggregation of remote MCP servers, a plugin registry, agents, skills and schedules.
+A four-layer memory system and Agent Skills compatibility, installed with a single Homebrew command.
03

Trade-offs

!It is macOS-only and built for Apple Silicon, and needs macOS 15.5 or later (some features need macOS 26).
!The scope is wide (runtime, agents, identity, memory, skills, schedules), so it is more than a simple local model server.
!Local model quality and speed are limited by your Mac's memory and chip.
05

Maintenance check

ActiveChecked: 2026-10-27
Last commit
2026-10-27
Commits, last 90 days
100+
Releases, last 90 days
30
Latest release
0.25.27 (2026-10-26)
Open issues
233
Recent issues with at least one comment
0/4
Commits per week, last 12 weeks
Commits per week, last 12 weeks: 59, 34, 74, 74, 60, 81, 63, 60, 47, 36, 58, 24

Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.

Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.

Most discussed open issues

  • #2519 Agent-loop premature stops, cache continuity, reasoning and MTP causal campaign30 comments · 2026-08-28
  • #823 I cant use tools, and I've granted all permissions28 comments · 2026-04-09
  • #647 no results whatsoever17 comments · 2026-03-15
  • #2211 Qwen3.6-35B-A3B-mxfp4 mtp 2.7 tok/s on power machine14 comments · 2026-07-28
  • #903 System prompts not injected at runtime14 comments · 2026-04-21

Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.

04

Who it fits

Pick it if
  • You use an Apple Silicon Mac and want local MLX models and cloud providers in one always-on app.
  • You want OpenAI-, Anthropic- and Ollama-compatible endpoints so existing tools can use it unchanged.
  • You want an MCP server, remote MCP aggregation, memory, skills and schedules in the same place.
Look elsewhere if
  • You are not on macOS 15.5 or later with Apple Silicon.
  • You only want a simple local model server: its scope is much wider.
  • You want a voice-first assistant: voice is not its focus.
06

How it differs

jarvis logovs jarvis★ 2.1k

Jarvis is a voice-first assistant for macOS, Windows and Linux. Osaurus is a macOS runtime and model server for apps and agents.

Full comparison →
07

Compatibility

macOS (Apple Silicon)
Platform
Verified via docs
Apple Foundation Models
Integration
Verified via docs
OpenAI API
API Standard
Verified via docs
Anthropic API
API Standard
Verified via docs
MCP Server
Protocol
Verified via docs
MLX Runtime
LLM Runtime
Verified via docs
08

Quick start

1
$ brew install --cask osaurus
09

Use cases

↳Automate complex multi-step tasks with AI agents using 'Work Mode' for code refactoring, debugging, or research.
↳Create personalized AI assistants with custom prompts, tools, and themes for various roles like Code Assistant or Creative Writer.
↳Automate recurring AI tasks and workflows such as daily journaling, weekly reports, or file organization via Watchers.
↳Integrate local and cloud Large Language Models for development and personal use via compatible API endpoints.
10

Alternatives

⚡Top VS:
VSjarvis★2.1k→VSOGAM★3.3k→VSrunanywhere-sdks★10.3k→VSharnessrouter★3.7k→
jarvis logo
jarvis★ 2.1k
Private local AI assistant for your desktop. Voice and text chat, personal memory, dictation and MCP tools. Runs offline with Ollama or OpenAI-compatible servers on macOS, Windows and Linux.
VSCompare→
OGAM logo
OGAM★ 3.3k
The Swiss Army Knife of Offline AI. Chat, see, speak, and generate images on your phone or Mac — GGUF LLMs, vision, Whisper speech-to-text, Stable Diffusion, tool calling, and local-network servers. Runs on your CPU, GPU, or NPU. No account, no API key, zero data leaves your device.
VSCompare→
runanywhere-sdks logo
runanywhere-sdks★ 10.3k
Production ready toolkit to run AI locally
VSCompare→
harnessrouter logo
harnessrouter★ 3.7k
HarnessRouter Community Edition: the self-hosted, Apache-2.0 edition of the unified interface for agent harnesses. Run Codex, Claude Code, Hermes, PI, DSH, and more through one API, with sessions, streaming, files, cancellation, and failure handling. Implements the Unified Harness Protocol (UHP), an open standard. Your keys, your infrastructure.
VSCompare→
trainingpeaks-mcp logo
trainingpeaks-mcp★ 187
TrainingPeaks MCP server for Claude Desktop, Code and Cowork. No API approval needed - works with any account. Query workouts, CTL/ATL/TSB fitness data, power PRs via natural language.
VSCompare→
houtini-lm logo
houtini-lm★ 121
Local or Cloud LLM support via MCP for your AI assistant with Houtini-LM - uses OpenAPI API for LM Studio, Cloud API and Ollama Compatibility. Save tokens by offloading some grunt work for your API - our tool description helps claude decide what work to assign and why.
VSCompare→
conduit logo
conduit★ 233
Local MCP gateway that cuts your agent's tool-token overhead ~90%. Every MCP server's tools collapse to 3 meta-tools the agent searches on demand. Measured: 97% less per request, same results. Keys in your OS keychain, no cloud.
VSCompare→
freellmpool logo
freellmpool★ 137
Free LLM API pool: 19 LLM providers cataloged, 235 routes, 355 cataloged chat models, keyless start when available.
VSCompare→
See all alternatives →

Featured in these lists

  • Best Open-Source Voice Assistants and Speech Tools

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs05Maintenance check04Who it fits06How it differs07Compatibility08Quick start09Use cases10Alternatives
osaurus-ai /
osaurus
↗

macOS runtime that runs local MLX and cloud models and exposes tools over MCP

8.1k483MIT
Last commit1d ago
StatusActive
CategoryLLM Infra
Contributors41
Latest release0.25.26
Links
Documentation↗Discussion↗Issues↗Releases↗
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit
© 2026 AgentIndex.app|Built by a 10-year iOS Developer.
Best by use caseAboutContactAdvertisePrivacy PolicyTerms of UseQYSGitHubBuy me a coffee ☕

Browse by Category

MCP ServersCode AssistantWorkflow AutomationRAG / Knowledge BaseMulti-AgentBrowser AutomationLLM InfraDev ToolingObservability

Not affiliated with Anthropic, OpenAI or Microsoft.