{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"hn-47124064","slug":"agent-multiplexer-manage-claude-code-via-tmux","name":"Agent Multiplexer – manage Claude Code via tmux","type":"agent","url":"https://github.com/mixpeek/amux","page_url":"https://unfragile.ai/agent-multiplexer-manage-claude-code-via-tmux","categories":["automation"],"tags":["hackernews","show-hn"],"pricing":{"model":"open_source","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"hn-47124064__cap_0","uri":"capability://automation.workflow.tmux.based.multi.agent.session.management","name":"tmux-based multi-agent session management","description":"Manages multiple Claude Code agent instances as isolated tmux sessions, enabling parallel execution and independent state management across agents. Each agent runs in its own tmux window/pane with separate environment variables, working directories, and execution contexts, coordinated through a central multiplexer process that routes commands and aggregates outputs.","intents":["Run multiple Claude Code agents in parallel without blocking on individual agent responses","Maintain isolated execution contexts for different coding tasks simultaneously","Switch between agent sessions without losing state or interrupting execution","Monitor multiple agents' progress and outputs in a unified terminal interface"],"best_for":["developers building multi-agent coding systems that need true parallelism","teams running distributed code generation tasks across multiple Claude instances","researchers experimenting with agent orchestration patterns"],"limitations":["tmux dependency adds operational complexity — requires tmux server running and proper session lifecycle management","No built-in persistence layer — agent state is ephemeral and lost on session termination unless explicitly serialized","Terminal-based output aggregation can become noisy with >5 concurrent agents without structured logging","Claude Code API rate limits apply per-agent, creating bottlenecks if agents exceed quota simultaneously"],"requires":["tmux 2.6+","Claude API key with Code Interpreter access","Python 3.8+ (inferred from GitHub repo structure)","Unix-like environment (Linux, macOS, WSL)"],"input_types":["natural language prompts","code snippets","file paths for context","structured task definitions"],"output_types":["terminal output streams","generated code artifacts","execution logs","structured JSON responses"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_1","uri":"capability://tool.use.integration.claude.code.api.command.routing.and.execution","name":"claude code api command routing and execution","description":"Routes user commands to Claude Code API endpoints through a multiplexed interface, handling authentication, request formatting, and response parsing. Translates high-level agent instructions into properly formatted API calls with context injection, managing the request-response cycle for code generation, execution, and artifact retrieval.","intents":["Send coding tasks to Claude Code and receive generated code without manual API call construction","Execute generated code and capture results within the agent workflow","Maintain conversation context across multiple API calls within a single agent session","Handle API errors and rate limiting gracefully without crashing the multiplexer"],"best_for":["developers building Claude-powered coding agents","teams automating code generation workflows","researchers prototyping multi-agent code synthesis systems"],"limitations":["API latency (typically 2-10 seconds per request) creates blocking points in agent execution","No built-in request batching — each command incurs separate API round-trip overhead","Context window limitations (200K tokens for Claude 3.5) constrain multi-file codebase awareness","API key management is manual — no built-in secret rotation or credential refresh"],"requires":["Valid Anthropic API key with Code Interpreter enabled","Network connectivity to api.anthropic.com","Python requests library or equivalent HTTP client"],"input_types":["natural language prompts","code snippets","file paths","execution parameters"],"output_types":["generated code","execution results","error messages","artifact metadata"],"categories":["tool-use-integration","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_2","uri":"capability://automation.workflow.agent.session.lifecycle.management","name":"agent session lifecycle management","description":"Manages creation, monitoring, and termination of individual agent sessions within tmux, including initialization with environment setup, health checking, and graceful shutdown. Tracks session state (active, idle, error, completed) and provides hooks for custom initialization and cleanup logic per agent.","intents":["Spawn new agent instances on-demand with isolated environments","Monitor agent health and detect hung or crashed sessions","Gracefully terminate agents and clean up resources","Persist agent metadata and execution history for auditing"],"best_for":["teams running long-lived multi-agent systems requiring reliability","developers building fault-tolerant agent orchestration","operations teams monitoring agent fleet health"],"limitations":["No automatic recovery — crashed agents require manual restart or external monitoring","Health checks are polling-based, adding latency to failure detection (typically 5-30 second detection window)","Session metadata is in-memory only — no distributed state store for multi-machine deployments","Cleanup is synchronous, blocking multiplexer shutdown if agents don't terminate within timeout"],"requires":["tmux 2.6+","Unix process management utilities (kill, ps)","Python 3.8+ with subprocess module"],"input_types":["agent configuration objects","environment variable maps","initialization scripts"],"output_types":["session IDs","status reports","health metrics","cleanup logs"],"categories":["automation-workflow","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_3","uri":"capability://automation.workflow.interactive.terminal.multiplexing.and.display","name":"interactive terminal multiplexing and display","description":"Provides a unified terminal interface for viewing and interacting with multiple agent sessions simultaneously using tmux's split-pane and window-switching capabilities. Aggregates output from parallel agents into a navigable terminal UI, allowing users to attach/detach from specific agent sessions or view all agents in a grid layout.","intents":["View output from multiple agents in a single terminal without switching windows","Send input to specific agents while monitoring others","Scroll through agent execution history and logs","Switch focus between agents using keyboard shortcuts"],"best_for":["developers debugging multi-agent systems interactively","teams monitoring agent execution in real-time","researchers observing agent behavior during development"],"limitations":["Terminal size constraints limit practical agent count to 4-6 visible panes before readability degrades","Output synchronization across panes can cause visual artifacts or lag with high-frequency logging","tmux configuration complexity increases with agent count — custom layouts required for >3 agents","No built-in filtering or search across agent outputs — requires external tools like tmux-grep"],"requires":["tmux 2.6+","Terminal emulator with 256-color support","Minimum terminal size 120x40 for reasonable multi-agent view"],"input_types":["agent session identifiers","layout configuration","keyboard input"],"output_types":["terminal display","captured pane content","session status indicators"],"categories":["automation-workflow","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_4","uri":"capability://automation.workflow.agent.command.queueing.and.execution.scheduling","name":"agent command queueing and execution scheduling","description":"Implements a command queue per agent that buffers incoming tasks and schedules execution based on agent availability and resource constraints. Prevents command loss during agent unavailability and enables priority-based execution ordering, with support for task dependencies and conditional execution based on prior results.","intents":["Queue multiple tasks for an agent without blocking on individual task completion","Prioritize urgent tasks over background work","Retry failed tasks automatically with exponential backoff","Execute tasks conditionally based on prior agent outputs"],"best_for":["teams running batch code generation jobs with varying priorities","developers building resilient agent workflows with retry logic","systems requiring task ordering and dependency management"],"limitations":["In-memory queue is lost on multiplexer restart — no persistent task store","No distributed queue support — single-machine only, doesn't scale to multi-node deployments","Priority scheduling adds complexity without guaranteed fairness — starvation possible for low-priority tasks","Dependency tracking is manual — no automatic DAG resolution or cycle detection"],"requires":["Python 3.8+ with queue module","Agent session must be active to process queued commands"],"input_types":["task objects with priority and dependencies","retry configuration","conditional execution rules"],"output_types":["task execution results","retry logs","queue status"],"categories":["automation-workflow","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_5","uri":"capability://data.processing.analysis.agent.output.capture.and.log.aggregation","name":"agent output capture and log aggregation","description":"Captures stdout/stderr from each agent session and aggregates logs into a centralized store with timestamps, agent identifiers, and severity levels. Implements circular buffering to prevent unbounded memory growth while maintaining searchable log history per agent and across all agents.","intents":["Access execution logs from completed agent sessions for debugging","Search across logs from multiple agents to find patterns or errors","Export agent logs for external analysis or compliance auditing","Monitor real-time log streams from active agents"],"best_for":["teams debugging multi-agent failures and tracing execution flow","systems requiring audit trails of agent activity","developers analyzing agent behavior patterns across runs"],"limitations":["Circular buffering limits historical log retention — older logs are discarded after buffer fills","No structured logging — logs are plain text, requiring regex parsing for extraction","Log aggregation is synchronous, adding latency to high-frequency logging scenarios","No built-in compression or archival — long-running agents can consume significant disk space"],"requires":["Disk space for log storage (configurable buffer size)","Python 3.8+ with logging module"],"input_types":["stdout/stderr streams from agent processes","log level configuration"],"output_types":["timestamped log entries","log files","search results"],"categories":["data-processing-analysis","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_6","uri":"capability://planning.reasoning.multi.agent.synchronization.and.coordination","name":"multi-agent synchronization and coordination","description":"Provides primitives for coordinating execution across multiple agents, including barriers (wait for all agents to reach checkpoint), message passing between agents, and shared state access. Implements distributed locking to prevent race conditions when agents access shared resources or coordinate on common tasks.","intents":["Wait for multiple agents to complete subtasks before proceeding to next phase","Pass results from one agent to another for sequential processing","Coordinate agents working on interdependent code generation tasks","Prevent race conditions when agents modify shared files or state"],"best_for":["teams building complex multi-agent workflows with dependencies","systems requiring coordinated code generation across multiple agents","developers implementing agent choreography patterns"],"limitations":["Synchronization primitives are in-memory only — no distributed locking across machines","Deadlock detection is manual — no automatic cycle detection in coordination graphs","Message passing is synchronous, blocking sender until receiver processes message","No timeout mechanisms on barriers — hung agents can block entire workflow indefinitely"],"requires":["Python 3.8+ with threading.Lock and queue.Queue","All agents running in same multiplexer instance"],"input_types":["barrier definitions","message objects","shared state keys"],"output_types":["synchronization events","message delivery confirmations","state snapshots"],"categories":["planning-reasoning","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"hn-47124064__cap_7","uri":"capability://automation.workflow.agent.configuration.and.environment.injection","name":"agent configuration and environment injection","description":"Manages per-agent configuration including environment variables, working directories, API keys, and custom initialization scripts. Injects configuration at session creation time, allowing agents to operate with isolated credentials and context without sharing state or secrets across agents.","intents":["Run agents with different API keys or credentials without cross-contamination","Set agent-specific working directories for isolated file operations","Inject custom environment variables for agent-specific behavior","Initialize agents with custom setup scripts or dependencies"],"best_for":["teams running agents with different credentials or configurations","systems requiring isolation between agent execution contexts","developers testing agent behavior under different configurations"],"limitations":["Configuration is static at session creation — no runtime reconfiguration without restart","No secret management — API keys stored in plaintext in configuration objects","Environment variable injection is shell-dependent — behavior varies across shells","No validation of configuration — invalid settings only detected at runtime"],"requires":["Python 3.8+ with os.environ","tmux 2.6+ for environment variable passing"],"input_types":["configuration dictionaries","environment variable maps","initialization scripts"],"output_types":["configured agent sessions","validation reports"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":34,"verified":false,"data_access_risk":"high","permissions":["tmux 2.6+","Claude API key with Code Interpreter access","Python 3.8+ (inferred from GitHub repo structure)","Unix-like environment (Linux, macOS, WSL)","Valid Anthropic API key with Code Interpreter enabled","Network connectivity to api.anthropic.com","Python requests library or equivalent HTTP client","Unix process management utilities (kill, ps)","Python 3.8+ with subprocess module","Terminal emulator with 256-color support"],"failure_modes":["tmux dependency adds operational complexity — requires tmux server running and proper session lifecycle management","No built-in persistence layer — agent state is ephemeral and lost on session termination unless explicitly serialized","Terminal-based output aggregation can become noisy with >5 concurrent agents without structured logging","Claude Code API rate limits apply per-agent, creating bottlenecks if agents exceed quota simultaneously","API latency (typically 2-10 seconds per request) creates blocking points in agent execution","No built-in request batching — each command incurs separate API round-trip overhead","Context window limitations (200K tokens for Claude 3.5) constrain multi-file codebase awareness","API key management is manual — no built-in secret rotation or credential refresh","No automatic recovery — crashed agents require manual restart or external monitoring","Health checks are polling-based, adding latency to failure detection (typically 5-30 second detection window)","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.36,"quality":0.26,"ecosystem":0.46,"match_graph":0.25,"freshness":0.6,"weights":{"adoption":0.25,"quality":0.25,"ecosystem":0.1,"match_graph":0.28,"freshness":0.12}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-06-17T09:51:04.692Z","last_scraped_at":"2026-05-04T08:09:59.925Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=agent-multiplexer-manage-claude-code-via-tmux","compare_url":"https://unfragile.ai/compare?artifact=agent-multiplexer-manage-claude-code-via-tmux"}},"signature":"+yXPSmUkfTI3ofU8+i/AP6XJ6UDdeMuEyjimX55q2oNn0Cq6nSnyZ0Z1X+d3WCix2za4/eBs30gHkPbWnUYcBQ==","signedAt":"2026-06-22T12:09:17.175Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/agent-multiplexer-manage-claude-code-via-tmux","artifact":"https://unfragile.ai/agent-multiplexer-manage-claude-code-via-tmux","verify":"https://unfragile.ai/api/v1/verify?slug=agent-multiplexer-manage-claude-code-via-tmux","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}