meridian
ModelFreeUse your Claude Max subscription with OpenCode, Pi, Droid, Aider, Crush, Cline. Proxy that bridges Anthropic's official SDK to enable Claude Max in third-party tools.
Capabilities14 decomposed
anthropic messages api protocol translation with session persistence
Medium confidenceTranslates stateless HTTP requests to the Anthropic Messages API (/v1/messages) into stateful Claude Code SDK query() calls, maintaining conversation continuity across proxy restarts through a fingerprint-based session cache that maps HTTP request lineage to SDK session objects. Uses session lineage verification to detect continuation, rollback, and compaction events, ensuring semantic correctness of multi-turn conversations without OAuth interception or binary patching.
Uses documented @anthropic-ai/claude-agent-sdk query() function with session lineage verification (message fingerprinting) to map stateless HTTP to stateful SDK sessions, avoiding OAuth interception or binary patching entirely. Implements session cache with fingerprint-based deduplication and rollback detection to handle conversation undo/compaction semantics correctly.
Unlike simple API proxies that forward requests unchanged, Meridian understands conversation semantics through lineage verification and can correctly handle conversation rollbacks and state compaction without losing SDK context.
openai chat completions api compatibility layer
Medium confidenceProvides OpenAI-compatible endpoints (/v1/chat/completions, /v1/models) that translate OpenAI request/response schemas to Anthropic Messages API format, enabling tools like Open WebUI or Continue that expect OpenAI-compatible APIs to work with Claude Max through Meridian. Handles model name mapping, message format conversion, and streaming response translation.
Implements bidirectional schema translation between OpenAI and Anthropic APIs at the HTTP layer, including message format conversion, model name mapping, and streaming response format adaptation. Maintains compatibility with OpenAI-first tools without requiring those tools to know about Anthropic.
Provides true OpenAI API compatibility rather than just accepting OpenAI-formatted requests; correctly translates response schemas and streaming formats so tools expecting OpenAI responses work seamlessly.
opencode plugin integration with custom header support
Medium confidenceProvides native integration with OpenCode IDE plugin, allowing OpenCode to use Meridian as a custom Claude Max provider. Implements OpenCode-specific header handling (x-meridian-profile, x-meridian-session-id) and response format adaptation. Includes plugin configuration examples and documentation for setting up OpenCode with Meridian.
Provides native OpenCode IDE integration with custom header support for profile switching and session management. Includes plugin configuration examples and documentation.
Unlike generic API proxies, Meridian's OpenCode integration understands OpenCode-specific requirements and provides seamless profile switching and session continuity.
extended context window management with model mapping
Medium confidenceMaps Claude model names to extended context window configurations, allowing agents to request specific context sizes (200K, 400K tokens) and automatically selecting the appropriate Claude model variant. Handles context window overflow by implementing sliding window or summarization strategies when conversation exceeds available context. Tracks token usage per request and warns when approaching context limits.
Implements model mapping to extended context window variants (200K, 400K) with automatic model selection and token usage tracking. Provides warnings when approaching context limits.
Unlike simple model proxying, Meridian's context management understands Claude's extended context variants and helps agents optimize for large codebases without manual model selection.
subagent routing and agent definition management
Medium confidenceSupports routing requests to subagents (nested agents) based on agent definitions and routing rules. Allows defining agent hierarchies where a parent agent can delegate tasks to specialized subagents. Manages subagent session isolation and result aggregation, enabling complex multi-agent workflows without requiring agents to know about each other.
Implements subagent routing with agent definition management, allowing parent agents to delegate to specialized subagents with session isolation and result aggregation.
Unlike flat agent architectures, Meridian's subagent routing enables hierarchical multi-agent systems where agents can delegate tasks without knowing about each other's implementation.
shared session store for multi-instance deployments
Medium confidenceProvides abstraction layer for session storage that supports both in-memory caching (default) and external stores (Redis, PostgreSQL) for multi-instance Meridian deployments. Implements session serialization/deserialization and distributed cache invalidation to ensure session consistency across proxy instances. Handles session expiration and cleanup policies.
Provides pluggable session storage abstraction supporting in-memory, Redis, and PostgreSQL backends with distributed cache invalidation for multi-instance deployments.
Unlike single-instance proxies, Meridian's shared session store enables horizontal scaling and high-availability deployments without losing conversation state.
multi-agent adapter detection and request normalization
Medium confidenceAutomatically detects which coding agent (OpenCode, Aider, Cline, Crush, Pi, Droid) is making a request through User-Agent analysis and working directory context, then applies agent-specific adapters that normalize tool definitions, file path formats, and working directory handling to a common internal representation. Each adapter implements the IAdapter interface to handle agent-specific quirks without modifying the core proxy logic.
Uses adapter-based architecture with automatic detection via User-Agent and working directory heuristics to support diverse agents (OpenCode, Aider, Cline, Crush, Pi, Droid) without requiring per-agent configuration. Each adapter implements IAdapter interface to handle agent-specific tool schema, file path, and working directory conventions.
Unlike single-agent proxies, Meridian's adapter system allows one proxy instance to serve multiple different agents simultaneously, each with their own tool definitions and path conventions, without manual configuration switching.
mcp tool system integration with dynamic tool registration
Medium confidenceIntegrates Model Context Protocol (MCP) tools into the Claude Code SDK's tool-use pipeline, allowing agents to call MCP-compatible tools (file operations, shell commands, web search) through the SDK's native tool-calling mechanism. Tools are registered dynamically via MCP server connections, and tool calls from the SDK are routed back to the appropriate MCP server with result streaming and error handling.
Bridges MCP tool servers into the Claude Code SDK's native tool-use pipeline, allowing agents to call MCP tools through documented SDK mechanisms rather than direct HTTP calls. Implements dynamic tool registration and result streaming with error handling.
Provides native MCP integration within the SDK's tool-calling flow rather than requiring agents to make separate MCP calls, resulting in tighter integration and better context preservation.
multi-profile account management with oauth token refresh
Medium confidenceManages multiple Claude accounts (profiles) within a single Meridian instance, allowing users to switch between profiles via HTTP headers (x-meridian-profile) or web UI. Automatically refreshes OAuth tokens on macOS and Linux using platform-specific credential storage, and routes requests to the correct profile's SDK session. Each profile maintains its own session cache and authentication state.
Implements multi-profile account management with automatic OAuth token refresh using platform-specific credential storage (macOS Keychain, Linux credential manager). Profiles are switchable at request-time via headers without proxy restart.
Unlike single-account proxies, Meridian supports multiple Claude accounts in one instance with automatic token refresh, eliminating manual credential management and enabling team-wide deployments.
intelligent session lineage verification and conversation rollback
Medium confidenceDetects conversation state changes (continuation, rollback, compaction) by computing message fingerprints and comparing request lineage to cached session state. When a rollback is detected (e.g., user undoes a step), the SDK session is rewound to the correct point in the conversation tree without losing context. Handles edge cases like message reordering and partial conversation history.
Uses message fingerprinting and lineage verification to detect conversation state changes (rollback, compaction) and automatically rewind SDK sessions to the correct point without losing context. Handles edge cases like message reordering and partial history.
Unlike simple session caching, Meridian's lineage verification understands conversation semantics and can correctly handle rollbacks and compaction, ensuring SDK state always matches the user's intended conversation flow.
streaming response handling with protocol-specific formatting
Medium confidenceHandles streaming responses from the Claude Code SDK and translates them into protocol-specific formats (Anthropic SSE, OpenAI SSE) with proper chunking, error handling, and connection management. Implements backpressure handling to prevent memory overflow when agents consume streams slowly, and supports both text and tool-use streaming with proper framing.
Translates Claude Code SDK's AsyncIterable streams into protocol-specific SSE formats (Anthropic and OpenAI) with backpressure handling and proper error recovery. Supports both text and tool-use streaming with correct framing for each protocol.
Unlike simple stream forwarding, Meridian's streaming layer handles protocol translation, backpressure, and error recovery, ensuring reliable streaming across different agent types and network conditions.
error handling and exponential backoff retry logic
Medium confidenceImplements comprehensive error handling for SDK failures, network errors, and rate limiting with exponential backoff retry logic. Distinguishes between retryable errors (rate limits, transient network failures) and non-retryable errors (authentication, invalid requests), and provides detailed error context to agents. Includes circuit breaker pattern to prevent cascading failures.
Implements intelligent error classification (retryable vs. non-retryable) with exponential backoff and circuit breaker pattern to handle transient failures gracefully. Distinguishes between rate limiting, network errors, and authentication failures.
Unlike simple retry loops, Meridian's error handling uses exponential backoff and circuit breaker to prevent cascading failures and provides detailed error context to agents for better debugging.
telemetry collection and monitoring dashboard
Medium confidenceCollects detailed telemetry on all proxy operations (request latency, token usage, error rates, streaming performance) and exposes a web-based monitoring dashboard for real-time visibility. Stores telemetry in a time-series database and provides API routes for querying historical data. Includes metrics for per-agent performance, per-profile usage, and SDK operation breakdown.
Provides built-in telemetry collection and web dashboard for monitoring proxy performance, token usage, and error rates across agents and profiles. Includes per-agent and per-profile metrics with historical data queries.
Unlike proxies without observability, Meridian includes a built-in monitoring dashboard and telemetry API, enabling teams to understand proxy behavior and optimize configuration without external tools.
docker deployment with environment-based configuration
Medium confidenceProvides Docker image and docker-compose configuration for easy deployment of Meridian in containerized environments. Supports environment variable-based configuration for profiles, MCP tools, telemetry storage, and proxy settings. Includes health check endpoints and graceful shutdown handling for Kubernetes and orchestration platforms.
Provides production-ready Docker image with environment-based configuration, health checks, and graceful shutdown for Kubernetes deployments. Includes docker-compose example for local development.
Unlike manual deployment, Meridian's Docker setup enables one-command deployment with proper health checks and orchestration support, reducing DevOps complexity.
Capabilities are decomposed by AI analysis. Each maps to specific user intents and improves with match feedback.
Related Artifactssharing capabilities
Artifacts that share capabilities with meridian, ranked by overlap. Discovered automatically through the match graph.
ollama
Get up and running with Kimi-K2.5, GLM-5, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
SGLang
Fast LLM/VLM serving — RadixAttention, prefix caching, structured output, automatic parallelism.
Ollama
Get up and running with large language models locally.
Llama 3 (8B, 70B)
Meta's Llama 3 — foundational LLM for instruction-following
Lepton AI
AI application platform — run models as APIs with auto GPU management and observability.
Free Models Router
The simplest way to get free inference. openrouter/free is a router that selects free models at random from the models available on OpenRouter. The router smartly filters for models that...
Best For
- ✓developers using third-party coding agents (Aider, Cline, OpenCode) with Claude Max subscriptions
- ✓teams deploying Meridian as a local proxy to avoid API key exposure
- ✓developers using OpenAI-first tools (Open WebUI, Continue IDE plugin) who have Claude Max subscriptions
- ✓teams wanting to provide Claude as a drop-in replacement for OpenAI in existing workflows
- ✓OpenCode IDE users with Claude Max subscriptions
- ✓teams using OpenCode as their primary coding agent
- ✓developers working with large codebases that need extended context
- ✓teams analyzing large documents or code repositories
Known Limitations
- ⚠Session cache is in-memory by default; requires external state store for multi-instance deployments
- ⚠Lineage verification adds ~50-100ms overhead per request for fingerprint computation and cache lookup
- ⚠Conversation state is tied to proxy instance; horizontal scaling requires shared session store implementation
- ⚠Model mapping is static; does not support dynamic model discovery from Anthropic
- ⚠Some OpenAI-specific parameters (e.g., logit_bias, top_logprobs) are silently dropped
- ⚠Streaming format translation adds ~20-30ms latency per chunk due to schema conversion
Requirements
Input / Output
UnfragileRank
UnfragileRank is computed from adoption signals, documentation quality, ecosystem connectivity, match graph feedback, and freshness. No artifact can pay for a higher rank.
Repository Details
Last commit: Apr 22, 2026
About
Use your Claude Max subscription with OpenCode, Pi, Droid, Aider, Crush, Cline. Proxy that bridges Anthropic's official SDK to enable Claude Max in third-party tools.
Categories
Alternatives to meridian
Are you the builder of meridian?
Claim this artifact to get a verified badge, access match analytics, see which intents users search for, and manage your listing.
Get the weekly brief
New tools, rising stars, and what's actually worth your time. No spam.
Data Sources
Looking for something else?
Search →