Power your agent
คำสั่งของผู้เผยแพร่เอง ไม่ใช่เอนด์พอยต์ของ PlugMCP
Host a copy with our keys and logs.
npx -y @dreb/[email protected]Claude Code ไม่มีลิงก์ติดตั้งสำหรับคำสั่งแบบโลคัล ให้คัดลอกคำสั่งด้านล่างแล้วรันในเทอร์มินัล
claude mcp add dreb-semantic-search -- npx -y @dreb/semantic-search@2.83.1รันในเทอร์มินัล
{ "mcpServers": { "dreb-semantic-search": { "command": "npx", "args": [ "-y", "@dreb/[email protected]" ] } }}claude_desktop_config.json
URL แบบรีโมตหรือคำสั่งบนเครื่องของคุณ? การติดตั้งสองแบบต่างกันอย่างไร
รันบน PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
เปิดจากผู้เผยแพร่
dreb is a hackable, open-source terminal coding agent and agent runtime for people who want to own their AI development workflow. It gives you a practical coding assistant today — tools, sessions, memory, model switching, subagents, and a polished TUI — while keeping the core flexible enough to reshape with skills, extensions, packages, custom providers, and alternate frontends. Its web dashboard puts every agent session in the browser: start work at your desk, steer it from your phone.
Semantic codebase search engine with embedding-based ranking and MCP server
Use dreb if you want a coding agent that can run against direct APIs, coding subscriptions, proxies, cloud providers, local models, or your own provider code; if you want workflows such as issue-to-merge automation and multi-agent review to be inspectable and replaceable; or if you want an agent runtime you can embed in a CLI, an RPC process, an SDK integration, a web dashboard, or a Telegram bot.
/login subscriptions, switch models at runtime with /model, scope model sets, and tune model-aware thinking levels through xhigh plus a model-aware max tier (offered only where the model supports them — e.g. Claude Opus 4.7+/Sonnet 5+/Haiku 5.5 send native Anthropic xhigh and max, Opus/Sonnet 4.6 send max only, and off is hidden for Claude Opus 5.5, whose thinking cannot be turned off). Codex ultra is orchestration (max plus local multi-agent work), not a raw provider effort. Route built-in providers through proxies, use cloud providers such as Bedrock/Vertex/Azure, or add local/proxy/custom models through Custom Models and Custom Providers. See Providers for the current setup list.subagent tool runs role-matched work in independent child agents using single, parallel, or chain mode. Omitting the agent type selects Explore, which retrieves concrete evidence such as files, symbols, documentation, call sites, exact snippets, and explicit data flows; the primary agent retains root-cause diagnosis, requirements interpretation, design, implementation recommendations, planning, synthesis, and final conclusions. Parallel and chain modes do not relax that boundary, while specialized agents continue to perform the broader work in their own definitions. Custom agent definitions can inherit models, record child-session metadata for audit trails, and power workflows such as mach6's specialized reviewers. Per-agent models and per-request thinking remain explicit controls. The optional singleModelMode setting instead forces every child onto the parent session's model—bypassing model overrides, per-agent model lists, and the dispatch arbiter—and prepends a warning to child output whenever a requested model selection was ignored. The built-in model-routing-guide skill researches scoped provider/model candidates and local child history, and its update mode preserves retained entries while removing stale scope and researching newly added models; the optional global-only Dispatch Arbiter consumes that guide in a fully headless, tool-less call before every child spawn and may change only agent, scoped canonical model, and supported thinking. Its bounded rolling parent activity follows the title setter and includes useful tool outputs, with existing secret scrubbing applied before inference. It is disabled by default and fails closed—bad configuration, guide, inference, or decisions prevent the child from spawning rather than silently keeping the original route. TUI /settings and dashboard Settings expose its enable toggle, exact model, thinking, guide path, and validation/readiness feedback. Typed decisions are persisted and visible in the TUI, JSON/RPC, and dashboard. The same settings surfaces expose backgroundAgents.maxConcurrentSubagents (default 4); 0 starts new parent sessions without the subagent tool and explicitly tells the parent model to perform normally delegated work itself. While background subagents run, a separate guardrail pauses the parent after a few turns, configurable via backgroundAgents./tree navigation, /fork, CLI --fork, compaction, HTML export, and JSONL import/export. Automatic compaction can run between tool-loop requests before the provider limit is reached; an opt-in continuation setting keeps pending or interrupted work moving after successful compaction without restarting an already completed answer or changing manual /compact. Memory is file-based, global + project-scoped, survives sessions, can read Claude Code project memory, and can be maintained with /dream memory consolidation.@, path completion, image paste/drag, bash shortcuts, hotkeys, settings, model cycling, steering/follow-up queues while the agent is working, token/cost/context status, custom themes, and extension-provided UI surfaces. Transcript prose, code, tool output, and agent results use terminal soft-wrap so copying from scrollback keeps long logical lines intact instead of injecting hard newlines./buddy hatches an Ollama-powered terminal companion with persistent state, generated personality/backstory, event reactions, idle quips, name-call responses, pet/reroll/stats commands, and a sidebar presence while you work.search. Semantic search uses AST-aware chunks, embeddings, POEM ranking, memory indexing, and also ships as @dreb/semantic-search with an MCP server for other harnesses. The semantic search package requires Node.js 22+.~/.dreb/agent/performance.jsonl). This data stays on your machine; the TUI footer, dashboard session details, Telegram /stats, and RPC share the same latest-100 median with long-term delta for personal analytics and model comparison.maxTokens is the ordinary request maximum, including model overrides; explicit bounded internal calls may use less. Genuine output-budget exhaustion fails loudly; possible full-context truncation enters one compact-and-retry recovery. The ChatGPT-backed Codex protocol is server-controlled because it rejects client output-limit fields.Node.js 22 LTS is required. dreb relies on SSE streaming behavior that is stable in Node 22 LTS. Node 24 and Node 26 are known to break provider streaming due to changes in ReadableStream buffering, which causes every provider to fail with "request ended without sending any chunks". If you see that error, switch to Node 22 LTS.
git clone https://github.com/aebrer/dreb.git
cd dreb
npm install
npm run build
npm link -w packages/coding-agent
See the full coding-agent docs in packages/coding-agent.
npm install -g @dreb/coding-agent
Authenticate with an API key and start the TUI:
export ANTHROPIC_API_KEY=sk-ant-...
dreb
Or use a coding subscription such as ChatGPT/Codex, GitHub Copilot, Gemini CLI, Antigravity, or Kimi:
dreb
/login
Or route through a custom provider — corporate proxy, OpenAI-compatible local server such as Ollama/LM Studio/vLLM, Bedrock proxy, Anthropic-compatible endpoint, Google-compatible endpoint, or extension-registered provider. See Custom Models and Providers.
Platform notes: Windows, Termux/Android, tmux, terminal setup, and shell aliases.
Bun users: Bun's lockfile can cache stale versions of @dreb/* packages, causing import errors after upgrades. If you hit missing export errors with bunx dreb, clear the cache and re-install:
bun pm cache rm
bunx --force dreb
dreb ships with 13 standard built-in tools enabled by default: read, write, edit, bash, grep, find, ls, web_search, web_fetch, subagent, wait, watch_github_ci (blocks until pull-request checks pass or fail), and ask_user (pause and ask the user one or more structured multiple-choice or free-text clarifying questions in a single batch wizard, with Markdown-formatted question text and an in-card Stop agent action, rendered natively in the TUI and Dashboard). Setting backgroundAgents.maxConcurrentSubagents to 0 removes subagent from new parent sessions. Three additional tools are always active: search for semantic codebase search, skill for loading workflows, and tasks_update for visible task tracking. suggest_next (ghost text command suggestions, Tab to accept) is active by default but excluded when --tools is specified.
Interactive mode adds slash commands such as /model, /settings, /resume, /tree, /fork, /compact, /dream, /buddy, /export, /reload, and /hotkeys. The message queue lets you steer a running agent or queue follow-up work without waiting for the current turn to finish.
GPT-6 models (Astra, Sol, Luna, 6.1 Sol) are available through ChatGPT/Codex and GitHub Copilot, with xhigh and native max thinking support. The Codex model list is derived automatically from OpenAI's catalog, so new models appear without a manual update (availability depends on your ChatGPT plan/workspace). Codex defaults to GPT-5.6 Luna; removed saved defaults produce a visible warning and prefer the same provider's available default before considering other providers. See provider setup and model-selection behavior.
dreb supports both subscription and API-key providers, with model metadata updated in releases. GitHub Copilot's model list is also refreshed from your account at login and token refresh, so models your plan doesn't offer are hidden. Current provider docs cover subscriptions such as Codex, GitHub Copilot, Gemini CLI, Antigravity, and Kimi; API-key providers such as Anthropic, OpenAI, Azure OpenAI, Google Gemini/Vertex, Amazon Bedrock, Mistral, Groq, Cerebras, xAI, OpenRouter, Vercel AI Gateway, ZAI, OpenCode, Hugging Face, Kimi, and MiniMax; plus custom local/proxy providers.
Custom model configuration can override built-in provider base URLs, merge custom models into built-in providers, set compatibility flags for OpenAI-compatible servers, resolve API keys from shell commands or environment variables, select Bearer-only auth for Anthropic-compatible endpoints, and register providers dynamically from extensions. A custom models[] entry or built-in modelOverrides entry in models.json can also set systemPrompt or appendSystemPrompt alongside its model metadata. The same behavior remains available through exact provider/model entries in settings.json modelSettings; configuring prompt behavior for one canonical model in both files fails loudly rather than applying implicit precedence.
When switching models, dreb replays exact-model signed, encrypted, or redacted reasoning state unchanged. Portable structured reasoning is limited to models on the same provider using the OpenAI Chat Completions API when the destination accepts the source's recognized plain field (reasoning_content, reasoning, or reasoning_text); other readable reasoning is carried as labelled plaintext, while opaque redacted or encrypted-only state is not sent to incompatible targets. Provider identity and signature compatibility matter for custom models as well. Switching changes only the outbound request, so returning to the original model can replay its original state unless history has been compacted or pruned.
Provider-specific docs include Kimi vision notes that distinguish the Kimi Code OAuth endpoint, the Kimi API-key coding provider, first-party Kimi CLI media handling, and Moonshot Open Platform vision support.
Skills provide progressively loaded instructions for specialized tasks. They can be invoked by users as /skill:name or by the model through the skill tool, support argument substitution, and can live globally, per-project, in packages, or on the CLI.
Extensions are TypeScript modules loaded with full access to dreb's extension API. They can add or override tools, intercept tool calls, mutate provider payloads, add commands and flags, define custom keyboard shortcuts, render custom tool output, open overlays and custom editors, persist state in sessions, register providers, surface warnings, and implement custom permission or workflow gates.
Resources carry source provenance so commands, tools, skills, and prompts can be traced through autocomplete, RPC discovery, and SDK introspection.
dreb packages make those resources installable and shareable through npm, git, URLs, or local paths. Use dreb install, dreb list, dreb update, and dreb config to manage them; project-local packages can be checked into settings so teams get the same skills, extensions, prompts, and themes.
Sessions are persistent JSONL files with a tree structure. You can resume recent sessions, browse past sessions, branch in-place with /tree, fork sessions into new files, compact long conversations, import/export JSONL, export HTML, or choose a custom session directory.
Main-session and subagent transcripts use separate roots. sessionDir (or CLI --session-dir) selects a flat main-session store; CLI precedence is flag, merged global/project settings, extension hook, then the built-in per-project tree. Dashboard host-wide inventory reads only the global sessionDir at startup, so restart Dashboard after changing it. Global-only subagentSessionDir controls new child logs; restart-time recovery also reads the legacy ~/.dreb/agent/subagent-sessions root when different. Absolute paths remain absolute, ~ expands from the home directory, and relative paths use the relevant CLI/Dashboard/AgentSession startup cwd. Changing either root does not migrate existing files or affect credentials or general configuration. See session settings.
Memory is just files. Global and project memory indexes are loaded into the system prompt at startup, and entries can store user preferences, good practices, project context, or navigation pointers. /dream backs up memory, merges duplicates, scans recent sessions for unrecorded patterns, prunes stale entries, and validates links.
At startup, dreb always loads project context files (AGENTS.md/CLAUDE.md) by walking upward from its launch cwd. This initial upward scan is separate from, and not controlled by, nested-context trust settings. Lazy nested/out-of-cwd loading is off by default: when a tool later enters a subdirectory or another repo, its context is loaded only if that canonical directory is covered by a global trusted root in ~/.dreb/agent/settings.json (context.trustedFolders). Trust a root only when you control its instructions; it covers that root and descendants after realpath resolution, so a symlink escaping the root is not trusted. In the dashboard, the Files view is the primary flow to grant or remove folder trust, while Settings lists every configured root for audit/revoke and offers add-by-path. context.autoLoadNested: true is a global-only expert trust-all override and can inject prompt-injection content from any resolvable directory. Project .dreb/settings.json cannot enable, disable, or extend nested-context trust, so a cloned repository cannot grant itself trust. Main agents and subagents use the same global decision. Auto-loaded content is secret-scrubbed and appended after extension tool_result transforms, which intentionally do not see it; paths are deduplicated so each file is injected at most once per session. See Context Files.
The same agent runtime powers multiple surfaces:
dreb -p for one-shot prompts, including piped stdin.@dreb/coding-agent and create agent sessions directly in TypeScript.@dreb/telegram runs dreb as a bot with sessions, model controls, file upload/download, live tool status, and visible results for user-facing tools.dreb dashboard serves a browser UI (fleet overview of all sessions, full chat with steering, subagent observability, host file browser, dreb memory editor); local-only by default, remote via Tailscale + rotating pairing code. See dashboard docs.The dashboard is the visual face of dreb: every agent session on the host, live in the browser, with the same fidelity as the terminal.
One host, every screen. The dashboard server and the TUI share the same sessions on disk and the same agent runtime. Start a refactor in the terminal, open the dashboard on your desktop to watch its subagents fan out, then pick the same session up from your phone on the couch — one synchronized state everywhere, streaming live over SSE. The layout is responsive by design: on a desktop it's a dense multi-column fleet; on a phone, fleet cards stack and their content wraps instead of overflowing, while the session view prioritizes read-and-steer because steering a running agent from wherever you are is the point. Its fleet sidebar collapses to a safe-area-aware overlay drawer on mobile — hidden by default, toggled from the session bar, closed by its close button, tapping outside, or Escape without stopping a pending agent question — so the transcript and composer keep the full width. The drawer supports keyboard focus navigation, and resizing the window across the mobile breakpoint preserves the separate desktop collapse and preferred-width settings without changing the mobile drawer width.
Fleet overview. Home base is every session across every project: live sessions with status chips (running / needs-attention / idle / error), activity lines, running subagents, task progress, context usage, and model — plus past sessions grouped by project, resumable with one tap. Sessions copied from another host or whose recorded project directory has moved remain visible; if that historical path is unavailable, Dashboard asks for an existing runtime project directory and keeps both paths visible without rewriting the session header. Terminal provider/API failures show their reason on the fleet card; transient failures clear that terminal card state when automatic retry begins. Initial load fetches live cards and opens SSE before the historical inventory scan finishes; unchanged session metadata is reused on later inventory and recovery reads. Live cards then update through compact SSE snapshots instead of re-fetching the full cross-project inventory on every turn, keeping weak mobile links responsive. When a session needs input, the browser tab badges and (opt-in) sends a service-worker notification (installable PWA, works on Android and iOS). For opt-in local mobile transport profiling, see the dashboard docs.
Full-parity session view. Not a reduced chat client: streaming markdown, thinking blocks, bespoke tool cards (read/write/edit/bash and markdown-rendering tools), sanitized inline PNG/JPEG/GIF/WebP images returned by any tool, task panels, queued-message chips, image attach/paste with sent-image transcript previews, built-in slash-command autocomplete and execution (including model/settings, scoped-models, session tree, fork, compact, import/export, dream, resume/reload, and new/quit), model/thinking switchers, fork-from-message, HTML export. A fleet sidebar beside the transcript lists every other live session with the fleet cards' project, status/reasons, activity/latest-assistant preview, subagent summaries, tasks, model, context, cost, and message/activity metadata. Cards stay deterministically ordered (project path, then session start time), and clicking one loads that session's transcript without carrying over the previous view's controls or drafts. Desktop collapse and draggable width persist browser-locally; the resize handle also supports arrow keys and Home/End. While hidden, the toggle border reflects the highest-priority other-session status (error → needs attention → running → idle). On phones it becomes an overlay drawer opened from the session bar. The mounted header stays synchronized with live runtime snapshots and authoritative detail refreshes; confirmed model/thinking changes do not flash back to older values, and compacted context usage remains visible as a conservative estimate until fresh provider usage arrives. /scoped-models opens the Settings editor for the session's current project context. Built-ins are intercepted generically and rejected fail-closed at the RPC prompt boundary, so unsupported or future commands show guidance instead of leaking into model input. Provider/API failures render inline on the failed assistant attempt with any partial output preserved, including after refresh or recovery; transient failures then switch the session status to retrying without erasing that history. Tool-result images remain visible to the human even when the active model is text-only. While the agent works you can steer (inject into the running turn), queue follow-ups, or stop — the same queue semantics as the TUI.
Notifications without surprise navigation. Session-scoped notices, warnings, and errors share one dismissible banner region inside the transcript column, so banners never move or shorten the fleet sidebar; long messages scroll within bounded space while controls remain reachable. The header groups back navigation on the left, session identity in the middle, live/details controls on the right, and model/thinking controls in a separate row. The fleet sidebar toggle sits on the left of the bottom usage/context-stats row, separate from the back link, and remains available when details are collapsed. App-global and other-session notices use a fixed top-center stack. Starting or closing a runtime never moves the browser automatically. A closed main-session or subagent page keeps its already-rendered transcript as a read-only browser snapshot with explicit Resume and Return-to-fleet actions until the user leaves, while Fleet simply removes the live card and keeps the on-disk session resumable in its project group.
Scoped models. Dashboard Settings includes an editor for the persistent model-cycling scope. It searches provider-grouped available models; offers model, provider, and all-model toggles; shows accessible up/down controls for the ordered partial scope; and has save/reset actions that work on mobile. An absent enabledModels value means implicit all models in registry order, including future registry additions, and that all-model view cannot be reordered. A saved partial scope is a non-empty ordered list of exact canonical provider/model references; editing a legacy glob, fuzzy, or thinking-suffix scope normalizes it to that form. The selected project context reads effective global plus project settings, but dashboard writes remain global and warn when a project value shadows the result. Changes seed new sessions only and never mutate running sessions. See Settings and the RPC settings contract.
Automatic tab titles. New unnamed sessions can be named automatically after a configurable number of tool calls. Dashboard Settings can enable or disable generation and pin its one-shot call to an exact provider/model or clear it back to the automatic route; when no dedicated model is set, dreb preserves the Explore-agent route and parent-session fallback. See Tab Title settings.
Low-data transcript images. Tool results and images uploaded with user turns remain visible through content-addressed browser references rather than full base64 in events and transcript JSON. In dashboard mode each unique image crosses the RPC stdout pipe at most once per child process — later occurrences are tiny references — and the child's 16 MiB stdout queue aborts only after 30 s without drain progress, so multi-image bursts no longer kill the session. The default browser-local mode lazily requests a preview bounded to 1024 × 1024 and 256 KiB; clicking enlarges that same preview without downloading the original. Settings also offers request-free placeholders and informed-opt-in automatic originals. Explicit originals disclose their size and confirm above 1 MiB. Authenticated same-origin routes accept only signature-matching PNG/JPEG/GIF/WebP, send nosniff, and recover evicted entries from authoritative session data; static GIF previews preserve animated originals behind the original route. Image bytes therefore cannot cause an SSE oversized-event resync. Full-resolution HTML exports remain self-contained and unchanged.
Live subagent observability and steering. Background subagents are first-class: the parent session has a bounded, scrollable panel listing every retained agent newest-first with full running/done counts, fleet cards show live counts, and each agent has a drill-in transcript that streams in real time and survives browser reloads. While a child is running, that view can queue the user's own steering messages directly to that child, showing its pending queue and effective one-at-a-time or all-at-once delivery mode; completed history remains read-only. The panel uses the task tracker's native collapse pattern and starts collapsed on mobile. When the Dispatch Arbiter is enabled, those views also show its host-validated changed, unchanged, or failed pre-spawn decision and the final routed agent/model/thinking—never raw arbiter output.
Live connection recovery. The top bar and persistent session header have an accessible, text-labelled connection indicator (connecting, connected, retrying, resyncing, disconnected, or auth failed). SSE uses bounded replay and an authoritative snapshot barrier to recover from browser reloads, server restarts, sequence gaps, slow-client backpressure, and stalls without duplicating state; task lists restore with the session snapshot. Named 25-second heartbeats and a foreground 60-second liveness watchdog detect a stuck stream. See dashboard recovery details and the RPC ordering contract.
Host files, explicitly. Browse the host filesystem, upload/download, create folders, and start a new session in any directory — every file operation logged server-side.
Memories, repairable. The Memories screen edits dreb memory scopes only: global ~/.dreb/memory plus populated .dreb/memory directories for active/disk project roots (empty projects are omitted because entry creation is outside this screen). It shows the complete MEMORY.md index (with a warning when it exceeds the 200-line prompt convention), opens local index links in the current scope, provides visible loading feedback, displays existing entry metadata or parse errors and sanitized Markdown previews, preserves drafts on exact-revision conflicts, and synchronously cleans matching index links before deleting an entry. It does not create/rename entries or expose Claude memory paths.
Curated appearance themes and fonts. A theme gallery in settings offers eight dashboard-native themes (entropist.ca, Dim, Solarized, Gruvbox, Caves of Qud, Van Gogh, plus the colorblind-safe Okabe-Ito and Paul Tol palettes), each with its own light and dark palette, plus system/light/dark mode and font selectors. The font picker offers 36 explicit choices plus Theme default: the existing Google-hosted IBM Plex Mono and bundled JetBrains Mono, Fira Code, Iosevka, OpenDyslexic (dyslexia-friendly), and Atkinson Hyperlegible (low-vision-friendly), plus 15 sans-serif and 15 serif families with real regular, italic, bold, and bold-italic styles. Visible options preview their own regular face in the open listbox; when closed, only the selected typography is used, so alternate self-hosted fonts remain lazy. Theme cards retain IBM Plex Mono for Theme default previews; the font picker's Theme default option previews the active theme's built-in family. Previews load as rows approach the visible scroll window rather than downloading every family at once. The additions are source-attributed Latin subsets with neutral internal names (all OFL-licensed). Restricted upstream names appear only in secondary “Based on …” descriptions; the derivative name is primary, and typeahead also searches source families. Existing IBM Google hosting is unchanged. No font uploads or new multilingual bundles are added; missing scripts use browser fallback. Choices are saved per browser and are independent of your TUI theme. See font provenance, picker controls and loading, and manual custom fonts in a private source build. There is no runtime font-upload or install-directory mechanism.
Launch locally:
dreb dashboard
# or: dreb-dashboard
Remote access is explicit and Tailscale-only. There is no LAN mode and no public exposure. Local mode binds loopback exclusively (with Host/Origin validation against DNS rebinding). Remote mode requires Tailscale — from your phone on the same WiFi or from the other side of the world, the path is identical:
dreb-dashboard --remote --allow [email protected]
Every remote request passes fail-closed layers: peer-specific tailnet identity resolution (concurrent requests for one peer share the in-flight lookup), identity allowlist (empty list denies everyone), first-login pairing with a rotating 6-digit code shown only on the host, then a signed device cookie. New pairings last 180 days by default; Settings can choose 1–3650 days for future pairings without changing existing expiry dates. A remote device in the final 10% of its recorded lifetime gets an advance warning at most once per UTC day. Paired devices and their expiry dates are listed in settings and can be unpaired at any time. Pairing grants terminal-equivalent power — the pairing screen says so before the code is entered.
Installable PWA + mobile notifications. The dashboard is an installable web app — add it to the home screen on Android Chrome or iOS Safari 16.4+ for a standalone, no-URL-bar experience. Needs-attention notifications go through the service worker (registration.showNotification) — the only path that works on Android (which removed the page Notification constructor) and on iOS (installed PWA only). For remote access from a phone, notifications and the service worker require a secure context (HTTPS), so the dashboard can terminate TLS itself using tailscale cert files — no reverse proxy, and the auth model is unchanged (the peer address stays the real tailnet IP):
dreb-dashboard --remote --allow [email protected] \
--https --cert /etc/dreb/cert.pem --key /etc/dreb/key.pem
Local mode (http://127.0.0.1) already qualifies as a secure context, so install and notifications work there with no TLS setup. See the dashboard docs for the one-time tailscale cert setup and renewal walkthrough.
For auto-restart on Linux, install a systemd user unit. Use the absolute path from which dreb-dashboard for ExecStart (the example below matches an npm global prefix under ~/.npm-global):
# ~/.config/systemd/user/dreb-dashboard.service
[Unit]
Description=dreb web dashboard
[Service]
ExecStart=%h/.npm-global/bin/dreb-dashboard
Restart=on-failure
[Install]
WantedBy=default.target
systemctl --user daemon-reload
systemctl --user enable --now dreb-dashboard
For auto-restart on macOS, create a LaunchAgent. Launchd runs with a minimal PATH, so invoke node directly on the resolved entry point (use command -v node and realpath "$(command -v dreb-dashboard)" — paths vary by install method; if realpath is not found, brew install coreutils):
Save as ~/Library/LaunchAgents/com.dreb.dashboard.plist:
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>Label</key>
<string>com.dreb.dashboard</string>
<key>ProgramArguments</key>
<array>
<string>/ABSOLUTE/PATH/TO/node</string>
<string>/ABSOLUTE/PATH/TO/@dreb/dashboard/dist/index.js</string>
<string>--port</string>
<string>5343</string>
</array>
<key>RunAtLoad</key>
<true/>
<key>KeepAlive</key>
<true/>
<key>ThrottleInterval</key>
<integer>10</integer>
<key>EnvironmentVariables</key>
<dict>
<key>PATH</key>
<string>/opt/homebrew/bin:/opt/homebrew/sbin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin</string>
</dict>
<key>StandardOutPath</key>
<string>/Users/YOU/Library/Logs/dreb-dashboard.out.log</string>
<key>StandardErrorPath</key>
<string>/Users/YOU/Library/Logs/dreb-dashboard.err.log</string>
</dict>
</plist>
# load / start
launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.dreb.dashboard.plist
# stop / unload
launchctl bootout gui/$(id -u)/com.dreb.dashboard
# check status
launchctl print gui/$(id -u)/com.dreb.dashboard | grep -E 'state|pid'
This is a LaunchAgent (must run as your user to read
~/.dreb/agentand spawn RPC children as you), not a LaunchDaemon. OAuth subscription creds in~/.dreb/agent/auth.jsonare found automatically viaHOME. If you use API keys via shell environment variables, add them toEnvironmentVariables— LaunchAgents do not source shell profiles. See the dashboard docs for full details.
WSL2 users: if you reach the dashboard from a Windows browser and hit an intermittent access-denied / pairing screen on http://127.0.0.1 after the WSL VM has been idle, see the WSL2 gotcha for the cause and keep-alive workarounds.
Full docs: dashboard.md.
dreb is a hard fork of pi-mono, itself derived from Claude Code. Claude Code is a great product; dreb is not trying to win by cloning every feature into a bigger built-in core. It is trying to win on control, hackability, provider choice, and inspectable workflows.
That means some features other tools bake in are intentionally left as user-space building blocks:
@dreb/semantic-search exposes an MCP server for other harnesses.The tradeoff is a smaller core with stronger escape hatches: markdown skills, TypeScript extensions, custom agents, custom providers, installable packages, and multiple frontends.
A hard fork means dreb controls the update cadence. Upstream changes do not land automatically; useful fixes can be cherry-picked, product direction can diverge, and dreb-specific work such as mach6, memory maintenance, Telegram, safety guards, and provider routing can evolve on its own schedule.
See FORK.md for details.
MIT
Semantic codebase search engine with embedding-based ranking and MCP server
{ "mcpServers": { "dreb-semantic-search": { "command": "npx", "args": [ "-y", "@dreb/[email protected]" ] } }}เครื่องมือ
ยังไม่มีรายการเครื่องมือ เซิร์ฟเวอร์แบบรีโมตจะได้รายการหลังการตรวจสอบสถานะครั้งแรก
FAQ
คัดลอกคอนฟิก stdio สำหรับ Cursor, Claude Desktop, Claude Code หรือ VS Code ได้ในหน้านี้ คอนฟิกนี้ใช้การติดตั้งของผู้เผยแพร่ ไม่ใช่พร็อกซีของ PlugMCP
ระเบียนนี้ไม่ได้ระบุตัวแปรสภาพแวดล้อมที่จำเป็น
@dreb/semantic Search ทำงานแบบโลคัลผ่าน stdio
ตรวจสอบล่าสุดเมื่อ 2026-11-15 Semantic codebase search engine with embedding-based ranking and MCP server