Connecter un agent
Ajoutez Busabase à l'agent que vous utilisez déjà
Chaque agent ci-dessous peut lire votre espace de travail et écrire via des demandes de modification, chacune avec un message et un diff. Choisissez le vôtre pour les commandes exactes, le fichier de configuration, et comment vérifier la connexion.
Lire le guide MCP complethttps://busabase.com/api/mcpWhat should your agent work with?
Choose the resource you need, then follow the setup guide for your agent below.
A database for agent work
Keep structured records with stable fields, sources, and version history. Use scoped credentials to decide which changes need review.
A knowledge base for your agent
Preserve reviewed findings beyond a conversation. Start with an agent-specific guide to records, sources, and reuse.
An apps library for agent tasks
Explore templates for organizing work around shared records. A template is a starting point; connections and permissions still need setup.
Agents en ligne de commande
Terminal AI pair programmer with deep git integration — but no MCP client.
Compétence agentD'après la documentation officielleSourcegraph's agentic coding tool — a terminal TUI with full shell, file editing and git access, plus editor extensions.
Compétence agentD'après la documentation officielleAnthropic's terminal coding agent, with user / project / local MCP scopes.
Compétence agentD'après la documentation officielleOpenAI's terminal coding agent, with a built-in MCP client and an /mcp panel.
Compétence agentD'après la documentation officielleTerminal coding agent with project-aware MCP configuration and support for remote HTTP servers.
Compétence agentD'après la documentation officielleOpen-source terminal coding agent from Charm with skills and built-in stdio, HTTP, and SSE MCP support.
Compétence agentD'après la documentation officielleLangChain's terminal coding agent built on Deep Agents, with planning, subagents, filesystem tools, and MCP configuration.
Compétence agentD'après la documentation officielleFactory's terminal agent runtime for interactive coding, headless execution, custom droids, skills, plugins, and MCP tools.
Compétence agentD'après la documentation officielleGoogle's open-source terminal agent; MCP servers go in settings.json under the httpUrl key.
Compétence agentD'après la documentation officielleGitHub's terminal coding agent (the `copilot` command) that reads, edits, and runs code in your local repo with your GitHub account and Copilot subscription.
Compétence agentD'après la documentation officielleBlock's open-source terminal agent, where MCP servers are called "extensions".
Compétence agentD'après la documentation officiellexAI's terminal coding agent (the `grok` command): an interactive fullscreen TUI with plan mode, parallel subagents in their own worktrees, headless mode (`grok -p`) for scripts, and ACP support for embedding in other apps. Launched in early beta on 2026-05-25 for SuperGrok and X Premium+ subscribers; it picks up a repo's AGENTS.md, plugins, hooks, skills and MCP servers out of the box and is fully compatible with Claude Code configuration with zero setup.
Compétence agentD'après la documentation officielleJetBrains coding agent available in the terminal, JetBrains IDEs, and headless CI/CD workflows.
Compétence agentD'après la documentation officielleMoonshot AI's terminal coding agent (TypeScript CLI, successor to kimi-cli) with MCP, Agent Skills, plugins and hooks.
Compétence agentD'après la documentation officielleXiaomi MiMo's terminal-native coding agent with persistent memory, skills, autonomous workflows, shell tools, and MCP connections.
Compétence agentD'après la documentation officielleMistral AI's terminal coding agent (the `vibe` command, package `mistral-vibe`): Vibe Code is the coding mode of Mistral's Vibe product, and one engine is shared across this CLI, a VS Code extension and Vibe Code Web sessions. It runs interactive or headless (`vibe -p`) sessions with built-in agent profiles, subagents, MCP servers and Agent Skills, against Mistral-hosted models or any OpenAI-compatible endpoint, including local models.
Compétence agentD'après la documentation officielleMeta's coding agent for large repos — plans changes, writes code, and validates results, fanning out to parallel sub-agents in isolated worktrees for big jobs. Runs from a CLI, powered by the Muse Spark model.
Compétence agentD'après la documentation officielleCoding-first Pi fork with built-in subagents, browser tooling, and native HTTP MCP discovery.
Compétence agentD'après la documentation officielleOpen-source terminal agent with dedicated MCP register/login subcommands and a built-in doctor probe.
Compétence agentD'après la documentation officielleOpen-source terminal coding agent with first-class remote MCP and OAuth support.
Compétence agentD'après la documentation officielleMinimal, extensible terminal agent harness with native Agent Skills discovery and shell tools.
Compétence agentD'après la documentation officielleOpen-source terminal coding agent from the Qwen team with shell tools and stdio or remote HTTP MCP support.
Compétence agentD'après la documentation officielleThe agentic terminal; MCP servers are added from settings, with no config file to edit.
Compétence agentD'après la documentation officielleExtensions IDE
Renamed: the Amazon Q Developer CLI is now the Kiro CLI. What remains is AWS's Amazon Q extension for VS Code, JetBrains, Visual Studio and Eclipse, supported until April 30, 2027, with new features shipping only in Kiro.
Connecteur MCPD'après la documentation officielleOpen-source autonomous coding extension for VS Code and JetBrains, with its own Remote Servers tab for MCP.
Compétence agentD'après la documentation officielleOpen-source IDE assistant configured from YAML blocks committed alongside the project.
Compétence agentD'après la documentation officielleCopilot agent mode in VS Code; MCP servers are declared per workspace or per user profile.
Compétence agentD'après la documentation officielleOpen-source VS Code coding agent that starts the OAuth 2.0 flow automatically for remote MCP servers.
Compétence agentD'après la documentation officielleAutonomous coding agent for VS Code with per-project and global MCP settings.
Compétence agentD'après la documentation officielleTencent's coding assistant: Settings → MCP on personal, Connector Management (MCP OAuth 2.1) on Enterprise.
Compétence agentD'après la documentation officielleAutonomous VS Code agent extension that can edit code, run terminal commands, and connect to local or remote MCP servers.
Compétence agentD'après la documentation officielleÉditeurs IA
The AI-first code editor; MCP servers are managed under Settings → Tools & MCP.
Compétence agentD'après la documentation officielleJetBrains' system for building software with agents. Air in IDEs runs coding agents such as Claude Agent, Codex, Gemini CLI, Junie and GitHub Copilot — plus any agent connected via the Agent Client Protocol (ACP) — side by side, with every change reviewable in context. Air Teams runs shared projects and automations in dedicated cloud environments, and Air Governance controls which agents, models and MCP servers an organization uses.
Compétence agentD'après la documentation officielleAWS's agentic coding tool: the Kiro IDE turns prompts into specs and working code, and the Kiro CLI, formerly the Amazon Q Developer CLI, brings the same agent to the terminal.
Compétence agentD'après la documentation officielleThe Cascade agent's editor; MCP servers live in a single JSON file and support OAuth per transport type.
Compétence agentD'après la documentation officielleHigh-performance collaborative editor; MCP servers are configured as `context_servers` and it runs the standard OAuth flow.
Compétence agentD'après la documentation officielleChat web
Web chat; add Busabase as a custom MCP server from ChatGPT Plugins (chatgpt.com/plugins), then @-mention it in Chat or Work.
Connecteur MCPD'après la documentation officielleWeb Claude (also Claude Desktop and Cowork): Customize → Connectors → Add custom connector, then paste the URL.
Connecteur MCPD'après la documentation officielleGoogle's Gemini app: connect a custom MCP server under Connected Apps and use it in chats and task threads. Personal Google Account, US, English only.
Connecteur MCPD'après la documentation officielleLindy is an AI employee / personal executive assistant you work with over iMessage, SMS, email and Slack; it connects to 1,500+ integrations and any hosted MCP server.
Connecteur MCPD'après la documentation officielleSlashy is an AI-native email and calendar client (Gmail, Google Workspace, Outlook) whose agent drafts replies, triages the inbox and acts on your behalf, and can be reached from iMessage and Slack.
Connecteur MCPD'après la documentation officielleByteDance's agent space, which accepts custom MCP services; it is also the outward extension surface for 豆包.
Connecteur MCPD'après la documentation officielleTencent's agent platform: Plugins → custom MCP plugin; also the outward extension surface for 元宝.
Connecteur MCPD'après la documentation officielleApplications de bureau
Agentastic.dev is a native macOS multi-agent IDE that runs coding-agent CLIs such as Claude Code, Codex, Gemini and Cursor side by side, each in its own git worktree, container or cloud VM.
Compétence agentD'après la documentation officielleAgentsRoom is a desktop app (macOS, Windows, Linux) with a mobile companion that runs many coding-agent CLIs, including Claude Code, Codex, Copilot CLI, OpenCode and Cursor, across your projects from one window.
Compétence agentD'après la documentation officielleZhipu's desktop general-purpose agent: the first one-click local OpenClaw app, with 50+ built-in skills, MCP services, IM channels (Feishu, WeCom, WeChat, DingTalk, QQ), goal mode and agent clusters.
Compétence agentD'après la documentation officielleAWS's AI assistant for work (launched as Amazon Quick, renamed AWS Quick on 2026-11-09): a chat agent on web, desktop and mobile that automates tasks, analyses data, builds web apps and runs research across connected files, email, calendar, CRM and databases. Free and Plus plans need no AWS account; an Enterprise subscription unlocks team connectors, including custom MCP integrations.
Connecteur MCPD'après la documentation officielleCodeAgentSwarm is a desktop Agentic Development Environment (macOS, Windows, Linux) that runs and supervises several coding-agent CLIs in parallel, including Claude Code, Codex CLI, Antigravity CLI, OpenCode and Cursor Agent.
Compétence agentD'après la documentation officielleMenlo Research's open-source, local-first desktop chat app that runs open models offline (and cloud models via your keys) and acts as an MCP host; a separate terminal 'Jan Agent' shares its MCP config.
Connecteur MCPD'après la documentation officielleMoonshot AI's general-purpose local agent for knowledge workers (Mac/Windows beta): Kimi Code as the kernel, skills, plugins, scheduled tasks and browser control.
Compétence agentD'après la documentation officielleMiniMax's desktop AI agent app (the former MiniMax Agent desktop, renamed) plus the `mcode` terminal CLI: local workspace, terminal, browser, skills, MCP, scheduled tasks, Telegram/WeChat/Lark/Feishu channels.
Compétence agentD'après la documentation officiellePaseo is an open-source agentic development environment (desktop, mobile, web and CLI) whose local daemon runs coding-agent CLIs such as Claude Code, Codex, Copilot, OpenCode and Pi in parallel worktrees.
Compétence agentD'après la documentation officielleA native (C++/Dear ImGui) desktop coding harness for Windows, macOS and Linux by Chris Wellons (skeeto), announced 2026-10-31. It connects to a model provider you supply (no providers built in), supports subagents, MCP servers, Agent Skills (SKILL.md) and desktop control, and uses around two orders of magnitude less memory than web-stack harnesses. Free Community Edition on GitHub; a paid Professional Edition adds built-in headless engines for remote deployment.
Compétence agentD'après la documentation officielleOpen-source (Apache-2.0) personal AI assistant from the AgentScope team: runs on your machine or cloud, with skills, plugins, MCP, scheduled tasks and DingTalk/Lark/WeChat/Discord/Telegram/iMessage/QQ channels. Renamed from CoPaw.
Compétence agentD'après la documentation officielleApplications mobiles
Pally is a free personal AI assistant you text on iMessage/RCS, WhatsApp, Telegram (SMS fallback); it connects to your apps and, per its changelog, to any MCP server you tell it about in chat.
Connecteur MCPD'après la documentation officiellePoke is a proactive personal AI assistant you text on Apple Messages, WhatsApp, Telegram and RCS; it connects to your email, calendar and other apps and can take custom MCP servers as integrations.
Connecteur MCPD'après la documentation officiellePlateformes d'agents
Open-source autonomous agent framework that runs in Docker and can use host terminal, files, and browser tools through A0 CLI.
Compétence agentD'après la documentation officielleBuda ships a built-in Busabase integration — connect it under Settings → Integrations, with no URL or config file to edit.
Compétence agentD'après la documentation officielleAnthropic's proactive agent app (iOS, Android and claude.ai), announced 2026-10-21 as a waitlist-only research preview for US Pro and Max subscribers. It works on its own cloud computer, sends a morning brief and asks before sensitive actions.
Connecteur MCPD'après la documentation officielleDeepSeek's open-source agent harness — a plugin-based framework (Cordis) with its own shell, file editing, and sandboxes, currently in developer preview.
Compétence agentD'après la documentation officielleCognition's autonomous software engineer: a cloud agent (app.devin.ai) that plans, writes, tests and reviews code in its own VM, plus a local Devin CLI. MCP servers, plugins and Agent Skills are managed on the Customize page. (The editor formerly called Windsurf is now Devin Desktop — a separate row.)
Compétence agentD'après la documentation officielleAlways-on AI teammates from xAI; each Bot gets a persistent cloud VM with a browser, filesystem, and terminal.
Compétence agentD'après la documentation officielleNous Research's autonomous terminal agent with persistent memory, skills, tool use, and local or remote MCP servers.
Compétence agentD'après la documentation officielleNEAR AI's open-source, security-first personal AI assistant (Rust): WASM-sandboxed tools, encrypted credential store, web UI, routines; also offered as a hosted deploy.
Compétence agentD'après la documentation officielleMoonshot AI's hosted OpenClaw: a dedicated cloud (or local desktop / Android) OpenClaw wired to Kimi models, chat channels and a ClawHub skill library.
Compétence agentD'après la documentation officielleOpen-source agent collaboration platform (formerly LobeChat; web, desktop and mobile) where agents with persistent memory take on tasks, work in team chats and run code locally or in a cloud sandbox. Skills use the open SKILL.md format and MCP servers are added under Settings → Connectors.
Compétence agentD'après la documentation officielleManus is an autonomous general AI agent that works in its own cloud sandbox (a virtual computer with internet access and a persistent file system) and takes actions instead of only answering.
Compétence agentD'après la documentation officielleMeta's consumer personal AI agent: each user shares a persistent cloud Linux VM (terminal, file system, browser) with the agent; chat in the Muse app, Mac app, or WhatsApp. US rollout; free tier plus paid plans.
Compétence agentD'après la documentation officielleHKUDS's ultra-lightweight, open-source, self-hosted personal AI agent (Python, PyPI `nanobot-ai`) with a WebUI, terminal UI, chat-app channels, skills and MCP.
Compétence agentD'après la documentation officielleNanoCo's open-source personal assistant that runs each agent in its own Linux container (Claude Agent SDK by default) and talks to you over WhatsApp, Telegram, Slack, Discord, iMessage and more.
Compétence agentD'après la documentation officielleNVIDIA's open-source (alpha) reference stack for running OpenClaw, Hermes or LangChain Deep Agents Code more safely inside NVIDIA OpenShell sandboxes with managed inference, network policy and managed MCP servers.
Connecteur MCPD'après la documentation officielleOpen-source software development agent platform with a local CLI, cloud runtime, shell tools, and configurable MCP servers.
Compétence agentD'après la documentation officielleSipeed's tiny open-source personal AI assistant in Go (runs on Raspberry Pi, RISC-V boards, Android) with channels, an exec tool, SKILL.md skills and MCP.
Compétence agentD'après la documentation officielleZeroClaw Labs' open-source personal AI agent: a single Rust binary that answers on 30+ channels, runs shell/browser/HTTP tools under a supervised-by-default policy, and acts as an MCP client.
Compétence agentD'après la documentation officielleZo is an always-on personal cloud computer (a Linux server with a shell, files, browser and 50+ tools) that an AI agent operates for you, reachable from chat, SMS, email, Telegram, Slack and Discord.
Compétence agentD'après la documentation officielleAn agent delivery platform: build agents in chat with Agent Builder, or run them through the Managed Agents API in an isolated cloud sandbox with shell, files, web tools, Skills and remote MCP servers.
Compétence agentD'après la documentation officielle