DeepBot Desktop Assistant
A desktop multi-agent assistant for enterprise workflows, scheduled work, persistent memory, and Feishu-connected communication.
- Source repo
- kevinluosl/deepbot
- Stars
- ★ 2.7k
- Last updated
- 5mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 23/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- An operations user wants one desktop assistant to handle routine file work, web research, document analysis, and outgoing notifications.
- Not a fit if
- Teams wanting turnkey Docker deployment; no image or Dockerfile is provided, only contacting the author
- Users relying on reasoning models such as o1 or DeepSeek-R1, which the README explicitly advises against
- Source review
- 23/100 · Major gaps 6 safety controls not found
What does this agent do, and when should you use it?
DeepBot is an Electron-based system-level AI assistant for personal productivity and enterprise workflow automation. Each UI tab maps to an isolated Session with its own Agent Runtime, context, and memory, while the Gateway handles session management, queues, routing, and cross-tab messages. Its main process includes agent-runtime, tools, scheduled-tasks, connectors, and database components, with a React renderer. It runs as a desktop application and documents macOS, Windows desktop, and Linux/Docker scenarios, although Docker access is described only as contacting the author. It supports Qwen, OpenAI, and Claude models, plus connectors for Feishu, WeChat, WeCom, and Smart KF.
DeepBot has the Gateway create and manage a Session for each tab. The Agent Runtime, built on @mariozechner/pi-agent-core, assembles instructions from AGENT.md, TOOLS.md, CUSTOM-TOOLS.md, MEMORY.md, memory-<tab-id>.md, and SKILL.md files before calling a model. Its built-in tools can read and write files, run shell commands, control a browser, fetch pages, search with Tavily, analyze documents, generate images, manage cron scheduled tasks, and exchange messages through Cross Tab Call. Connectors receive and send text, images, and files through Feishu, WeChat, WeCom, and Smart KF; the Feishu document tool can create, read, append, update, and delete blocks and add comments. File and command operations pass a path-whitelist check covering the configured workspace, scripts, skills, and generated-image directories.
- An operations user wants one desktop assistant to handle routine file work, web research, document analysis, and outgoing notifications.
- A Feishu-based team needs bot DMs or @mentions in group chats to create separate sessions and send users text, images, or files.
- A project lead wants separate sales, marketing, product, and engineering tabs with independent memories that can collaborate through Cross Tab Call.
- A user needs a recurring daily check or report and wants to create a cron-based Scheduled Task with execution history.
- A TypeScript developer needs to add an internal business tool through the ToolPlugin interface in src/main/tools/.
How do you install or deploy this agent?
Requirements are Python 3.11+, Node.js 20.0.0+, and pnpm 10.23.0+. Run: git clone https://github.com/kevinluosl/deepbot.git; cd deepbot; pnpm install; pnpm run dev. To use a model, configure an API key for the chosen Qwen, OpenAI, or Claude provider in System Settings; the repository does not document provider-specific key fields or endpoints. For Feishu, open “System Settings” → “External Connections” → “Feishu” and enter an App ID, App Secret, and Bot Name.
How do you use this agent?
After startup, each tab is an isolated Session. Enter a task in the app or create a scheduled task; the documented example is “Check desktop files every day at 9 AM.” Use Cross Tab Call when sessions need to collaborate. Before using Feishu, save the connector credentials and start the Feishu connector, then configure the documented DM and group policies as needed. Build desktop packages with pnpm run dist, or use pnpm run dist:mac:local for an unsigned local macOS build.
What are this agent's strengths and limitations?
- Tabs isolate sessions, memories, and contexts while Cross Tab Call enables explicit collaboration across roles.
- Its documented tool surface spans files, shell commands, browser control, web fetching, document analysis, image generation, scheduled tasks, and several messaging channels.
- System prompts are dynamically assembled from base prompts, tool instructions, global and per-tab memory, and skill instructions, with documented hot updates.
- A configurable path whitelist validates file and command actions across the workspace, scripts, skills, and generated-image directories.
- Shell execution and filesystem access are core capabilities; the default workspace is the user home directory, so whitelist settings require deployment review and tightening.
- Model access requires user-provided Qwen, OpenAI, or Claude API keys; the repository does not document pricing, provider-specific behavior, or service-failure fallback.
- Linux Docker deployment is mentioned but lacks a copyable image, Dockerfile, or deployment command; users are told to contact the author.
- The README advises against explicit reasoning-mode models and notes that unsigned local macOS builds trigger security warnings.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| DeepBot Desktop Assistant This agent | 23 · Major gaps | Desktop appFree + model costs | ★ 2.7k | 5mo ago | TypeScript | OpenAI API · Claude API |
| holaOS | 52 · Major gaps | Desktop appFreemium | ★ 11k | 9d ago | TypeScript | Codex · Claude Code · OpenAI API · Claude API |
| ClawX Desktop | 45 · Major gaps | Desktop appFree + model costs | ★ 7.6k | 1d ago | TypeScript | Codex · OpenAI API |
| Open Cowork | 65 · Some gaps | Desktop appFree + model costs | ★ 2.3k | 17d ago | TypeScript | Claude Code · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows a path whitelist mechanism (least_privilege=1), but no implementation details for user confirmation or data flow transparency. Sensitive data handling and dependency security are not mentioned. External effects and rollback mechanisms are not described. Source attribution is partially evident through author field and MIT license (source_attribution=1).
Self-consistency: README claims 20+ tools but architecture diagram shows 14, inconsistency (self_consistency=1). Dependency availability not verified, failure messages not mentioned.
Audience and scenarios are clear (enterprise productivity), but capability boundaries are vague (capability_boundaries=1). Trigger precision not specified, environment fit partially covered (macOS/Windows/Linux) but not in depth (environment_fit=1).
Information architecture is clear (information_architecture=2), install notes are detailed (install_notes=2), naming stability partial (naming_stability=1), examples and FAQ provided (examples_and_faq=2), known limitations not mentioned, license is clear (license=2), versioning and changelog missing, maintenance responsibility partially evident through author info (maintenance_responsibility=1).
Output usability not specified, marginal value described but not quantified, cost-benefit not analyzed.
Claim traceability partial (claim_traceability=1), but lacks cross-source corroboration and fact-inference separation.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- README claims 20+ tools but architecture diagram shows 14, inconsistency.
- No user confirmation mechanism, potentially automatic execution of sensitive operations.
- Dependency security not mentioned, potential risks.
- No rollback mechanism, operations may be irreversible.
- No known limitations, users may misuse.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
1. 1. Skill 能力优化 4. 支持 Docker 模式 7. 支持微信、飞书、企业微信、智能客服 10. 记忆优化，节省 Token 13. 浅色主题与体验优化 访问https://glint-mvt.com/deepbot/获取更多信息View on GitHub ↗