meridian
Use Claude and Antigravity with Pi, OpenCode and other coding clients. Local API bridge, usage dashboard and Mac app. Antigravity preview available in 1.74.0.
Install / Use
npx skills add rynfar/meridianInstalls into whichever agent you are using.
Aider Config
Aider AI pair programming config
Quality Score
Category
AI & Machine LearningSupported Platforms
Our assessment of meridian
meridian scores 91/100 on our quality scale, 298th of 965 AI & Machine Learning skills we index (top 31%).
Its Aider Config is 8.1 KB long, well organised into 8 sections with 2 code examples: a thorough specification that gives an agent plenty to work with.
With 2,178 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated today, so meridian is actively maintained.
- Our last check on 2026-10-19 found the source still online.
- No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
- Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-09-24. Automated pattern scan on 2026-09-24. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
meridian compared with similar skills
All 4 of these similar skills score higher than meridian; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| meridian (this skill)by rynfar | 91 | 2.2k | today | Aider Config |
| claude-memby thedotmack | 100 | 113.5k | today | CLAUDE.md |
| Agent-Reachby Panniantong | 100 | 109.2k | today | CLAUDE.md |
| Understand-Anythingby Egonex-AI | 100 | 90.9k | 10d ago | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 78.9k | today | CLAUDE.md |
Frequently asked questions
- How do I install meridian?
- Run
npx skills add rynfar/meridian. The install tabs above show the steps for each supported agent. - Which AI agents does meridian work with?
- It is written for Aider, Cline and Gemini CLI, as a Aider Config file. Other agents that read the same format can often use it too.
- Is meridian safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is meridian still maintained?
- The repository was last updated today, so meridian is actively maintained.
Skill content
View source on GitHubMeridian connects coding clients to Claude and Antigravity using your subscription. Run Pi, OpenCode or another supported client against a local API, and manage providers, requests and usage from a browser or the optional Mac app.
Claude requests run through Anthropic’s Agent SDK. Antigravity requests run
through Google’s official, signed-in agy CLI. Account permissions, model
availability and subscription limits still apply.
| <img src="assets/providers/claude.png" width="40" alt="Claude logo" /><br>Claude | <img src="assets/providers/antigravity.png" width="40" alt="Antigravity logo" /><br>Antigravity |
| :--- | :--- |
| Released · Claude Agent SDK | Preview · v1.74.0+ · Official agy CLI |
| Persistent sessions, prompt caching and multiple account profiles | Streaming, client tools, conversation reuse and saved-answer recovery |
| Client compatibility and setup | Pi and OpenCode setup |
| Headless macOS, Linux and Windows | Authenticated acceptance on macOS arm64; Linux/Windows verification unfinished |
Quick start
For the released Claude backend, install Node.js 22+, then:
npm install -g @rynfar/meridian
claude login
meridian
Open http://127.0.0.1:3456. For OpenCode V1, run meridian setup and restart
OpenCode. For Pi and other clients, follow the client setup guide.
See configuration for API-key protection and Windows
setup, or deployment for Docker and Nix.
Try Antigravity
Antigravity is included in Meridian 1.74.0 and later as a preview. The verified setup uses agy 1.2.7, Pi 0.72.1 or OpenCode V1 1.18.31, and macOS on Apple Silicon. Other versions need their own verification.
Install Node.js 22+ and the
official Antigravity CLI. Run agy to sign in with
your subscription account and disable paid overage credits in its settings.
Then install Meridian:
npm install -g @rynfar/meridian@1.74.0
MERIDIAN_BACKEND=antigravity \
MERIDIAN_AGY_ALLOW_TOOL_BRIDGE=1 \
MERIDIAN_AGY_STATE_PATH="$HOME/.local/state/meridian/antigravity.sqlite" \
MERIDIAN_PORT=3457 \
meridian
Open http://127.0.0.1:3457/providers and use Connect Pi / OpenCode, or follow
the CLI setup instructions.
For a source build, follow the development guide.
The tool bridge lets the client execute tools with its own permissions. The
SQLite path enables saved state across service restarts; it does not guarantee
recovery of an unfinished action.
What works
- The coding loop: incremental text, client tools, parallel tool batches, approvals, questions, cancellation and client-owned plugins/delegation.
- Model and output selection: available model/effort variants, forced tools and validated structured output.
- Client APIs: Anthropic Messages, OpenAI Chat Completions and a documented Responses subset.
- Attachments: images and public HTTPS image URLs; documents, speech and video through local conversion. These are not native media APIs.
- Continuity: warm conversation reuse, eligible completed-session restoration, history replay and bounded recovery of saved answers and tool IDs.
- Provider management: web and macOS setup, separate quota/status displays, shared navigation and activity. Native browser/subagent access requires separate explicit grants.
What is unfinished
Active-task reattachment and automatic reconciliation of uncertain tool outcomes are unfinished. Some provider-specific plugins, hosted-tool/file contracts and client versions need adapters and acceptance tests. Authenticated Linux/Windows verification is also outstanding.
The CLI does not expose exact output-token caps, numeric thinking budgets, sampling controls or every native media/reasoning feature. Meridian cannot add those guarantees through a wrapper. Upstream CLI failures can still interrupt a turn. The support and recovery checklist records what is missing, available workarounds and what would close each gap.
Desktop
<p align="center"> <img src="assets/desktop-dashboard.jpg" alt="Meridian Desktop showing usage limits, cache activity and recent requests. Sample data." width="1000" /> </p> <p align="center"><sub>macOS dashboard preview · Sample data</sub></p>The optional Mac app manages the local service, versions, plugins and diagnostics. Use Providers to navigate Claude and Antigravity; their accounts and quota windows stay separate. Request history helps track failures and activity across the service.
The desktop preview targets macOS on Apple Silicon. Download the signed, notarized DMG or ZIP from GitHub Releases, or follow the source instructions. To try Antigravity, connect the app to a Meridian 1.74.0+ service configured as shown above. Older Meridian versions do not include Antigravity.
Headless use remains supported. The npm package does not install Electron; Docker, Nix and existing service managers can run the same server.
How it works
<p align="center"> <img src="assets/how-it-works.svg" alt="Pi, OpenCode and other supported clients connect to Meridian. Claude uses the Agent SDK; Antigravity uses the official agy CLI. Each provider uses its own signed-in account." width="920" /> </p>Meridian translates supported client requests at the provider boundary. It keeps Claude-specific account routing and cache behavior separate from Antigravity’s CLI sessions and permissions. Antigravity generation stays on the signed-in CLI; there is no Google API-key or model-SDK fallback.
Documentation
| Guide | Contents | | --- | --- | | Antigravity setup | Models, client tools, attachments, permissions and setup | | Antigravity support & recovery | Feature status, unfinished work and recovery limits | | Claude client setup | Compatibility matrix and client configuration | | Desktop | Local preview, service ownership and platform status | | Configuration | CLI, environment variables and API-key protection | | Accounts & profiles | Claude sign-in, multiple accounts and routing | | Deployment | Docker, Nix and headless services | | Plugins | Official packages and plugin configuration | | Monitoring | Usage, request diagnostics and prompt caching | | Development | Build, test and programmatic API | | Live verification | Tested versions, flows and known failures |
Contributing
Report an issue, open a PR or join
Discord. Read AGENTS.md,
ARCHITECTURE.md and E2E.md before changing behavior.
For Antigravity reports, include your OS, agy and client versions, the failing
flow and redacted diagnostics.
Meridian is MIT licensed. Claude and Antigravity are trademarks of their respective owners. Provider asset credits.
Related Skills
claude-mem
113.5kPersistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Agent-Reach
109.2kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
Understand-Anything
90.9kGraphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
headroom
78.9kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
