Getting Started

lm-assist

The Observability Platform for Claude Code & Agent SDK — Monitor, debug, and control AI coding agents with full session visibility, real-time execution tracking, and a complete REST API.

lm-assist provides the observability layer that Claude Code and the Agent SDK don't have. Every session, every subagent, every tool call, every token spent — visible through a web dashboard, a 155+ endpoint REST API, and SSE event streams. Access from localhost, your phone on the same WiFi, or anywhere via langmart.ai.

The Problem

Claude Code and the Agent SDK produce rich session data but provide no dashboard, no monitoring, and no debugging tools. You get a terminal or logs. When you need to understand what an agent did, why it failed, or how much it cost — you're on your own.

Key Capabilities

Capability What It Does
Monitor Real-time execution tracking — live session status, running process detection, per-model cost & token breakdown, SSE event stream
Debug 15 insight views per session — Chat, Thinking, Agents, Plans, Team, DAG, Files, Git, Console, Skills, Commands, Tasks, Summary, Meta, JSON
Control Web terminal from any browser, start/abort agents via API, SDK runner for headless execution, remote access via cloud
Session API 27 endpoints for session data — list, detail, delta fetch, batch-check, conversation, subagents, forks, DAG
Multi-Machine Connect a fleet of dev machines via LangMart Hub — browse sessions across all hosts, unified cost tracking
Knowledge Optional subsystem — auto-extract knowledge from sessions, inject via MCP tools, enable/disable at runtime to save memory

Data Sources

lm-assist reads the same JSONL session files regardless of how they were created:

Source What It Produces
Claude Code CLI Interactive sessions with subagents, teams, worktrees
Agent SDK (Python/TypeScript) Programmatic agent executions — same JSONL format
Headless mode (claude -p) Background/CI runs
Running processes Live PID, tmux, terminal state (real-time)

Quick Start

In Claude Code, run:

/plugin marketplace add langmartai/lm-assist

/plugin install lm-assist@langmartai

Then open a new Claude Code session and run:

/assist-setup

See Installation for alternative install methods (npm, from source).

Screenshots

Session Browser

Session Browser

Browse all sessions with human-readable names, live status, per-model cost, and token usage.

Session Detail (15 Views)

Session Detail

Full conversation timeline with syntax-highlighted code blocks — one of 15 specialized views.

Agent Tree

Agent Tree

Visualize subagent hierarchy — Explore, Plan, Bash, and custom agents with status tracking.

Task Kanban

Task Kanban

Aggregated task board from all active sessions with dependency tracking.

Web Terminal

Web Terminal

Live terminal access to running Claude Code sessions from any browser.

Architecture

DATA SOURCES                           lm-assist
                                       ┌─────────────────────────────────────┐
  Claude Code ──────────┐              │  Session Engine (Core API :3100)    │
  (interactive CLI)     │              │                                     │
                        ├──  JSONL ──▶ │  JSONL Parser → Session Cache       │
  Agent SDK ────────────┤   sessions   │  → DAG Builder → Cost Calculator    │
  (Python/TypeScript)   │              │  → SSE Streaming → SDK Runner       │
                        │              │                                     │
  Headless / CI ────────┘              └──────────┬──────────────────────────┘
                                                  │
  Running Processes ──── PID/tmux ──▶             │
                                       ┌──────────┴──────────────────────────┐
                                       │  Three Pillars                      │
                                       │                                     │
                                       │  MONITOR    DEBUG      CONTROL      │
                                       │  Sessions   15 views   Web terminal │
                                       │  Cost       DAG        Abort/start  │
                                       │  SSE        Subagents  SDK runner   │
                                       │  Fleet      Forks      Hub relay    │
                                       │                                     │
                                       └──────────┬──────────────────────────┘
                                                  │
                                       ┌──────────┴──────────────────────────┐
                                       │  Web UI (:3848)                     │
                                       │  15 tabs · responsive · real-time   │
                                       │                                     │
                                       │  Access: localhost · LAN · cloud    │
                                       └─────────────────────────────────────┘

Access Modes

lm-assist operates in three access modes simultaneously. Your data always stays on your machine.

Feature Local LAN Cloud
Dashboard access localhost only Same network Anywhere
Terminal access Same machine Same network Any browser, anywhere
Authentication None Optional token API key required
Data storage On machine On machine On machine (relayed, not stored)
Multi-machine No No Yes

Local Mode — http://localhost:3848. No auth. Default.

LAN Mode — http://192.168.x.x:3848 from phone/tablet. Optional LAN auth token.

Cloud Mode — https://langmart.ai/assist-dashboard. WebSocket relay through Hub. Get an API key at langmart.ai/assist.

Key API Surface

Category Endpoints Highlights
Sessions 27 List, detail, delta, batch-check, conversation, subagents, forks, DAG
Monitor 6 Running executions, summary, abort, SSE stream
Projects 12 List projects, sessions per project, git info, worktree detection
Terminal 13 ttyd start/stop/status, WebSocket proxy, tmux attach
Tasks 22 Task lists, aggregated tasks, ready tasks, dependency tracking
Knowledge 21 Optional — list, search, generate, review

Who It's For

  • Solo developers using Claude Code — see what's happening across all your sessions
  • Teams building with Agent SDK — observability for your agent pipelines
  • DevOps managing agent fleets — multi-machine dashboard, cost tracking, process management
  • AI product builders — debug agent behavior with 15 insight views

Next Steps