Cross-platform chatbot framework made with love
Full LLM thinking from the 4-phase benchmark pipeline.
{
"service_type": "platform",
"base_url": "https://koishi.chat",
"auth_method": "none",
"auth_config": {},
"endpoints": [],
"pricing_model": {
"type": "free",
"details": {
"note": "Koishi is an open-source framework. No public pricing model or paid tiers are documented; usage costs are tied to the underlying self-hosted infrastructure and third-party messaging platform requirements."
}
},
"rate_limits": {},
"capabilities": [
"Cross-platform chatbot framework",
"Extensible plugin system",
"High-performance message handling",
"Adapters for Discord, Telegram, Slack, Lark, DingTalk, QQ, Line, Matrix, WhatsApp, WeCom, WeChat Official, Kook, Zulip, Mail, Satori",
"Database backends: Memory, MongoDB, MySQL, PostgreSQL, SQLite",
"Built-in web console / admin dashboard",
"Plugin marketplace and market management",
"Command system and middleware pipeline",
"Internationalization (i18n) and localization",
"Schema-based configuration system",
"Session, event, and context APIs for bot development",
"Server and HTTP service integration",
"Message rendering and element system",
"Permission management (experimental)",
"Resource manager (Explorer)",
"Logging, analytics, and status monitoring",
"Sandbox debugging",
"Docker, Linux, macOS, Windows, and Android installation support",
"CLI launcher and template project generation",
"Plugin development, publishing, and testing tooling",
"Satori protocol support"
],
"raw_analysis": "Koishi (https://koishi.chat) is an open-source, cross-platform chatbot development framework. It is aimed at developers who want to build extensible, high-performance chat bots that can run across many messaging platforms. The documentation is extensive and mature, organized into manuals, development guides, API references, schema docs, plugin docs, and release notes (up to v4.15), indicating a long-lived, actively maintained project with a stable v4 line.\n\nThe framework's core value is its adapter/plugin architecture. Documented official adapters cover Discord, Telegram, Slack, Lark, DingTalk, QQ, Line, Matrix, WhatsApp, WeCom, WeChat Official, Kook, Zulip, Mail, and Satori, which makes it effectively a multi-platform integration layer for bots. It also supports multiple database backends (memory, MongoDB, MySQL, PostgreSQL, SQLite), a web console with plugin market, command management, permissions, logging, analytics, and a sandbox. There is an experimental theme/console extension system, resource explorer, and permission management.\n\nKoishi itself is a framework and runtime rather than a hosted SaaS platform. Users install it on Linux, macOS, Windows, Android, or via Docker, or use it as a dependency. There is a CLI launcher and template project generator. Because it is a local/self-hosted framework, there is no public REST API, no authentication scheme for an official hosted service, and no documented rate limits. Pricing is effectively free/open-source; any costs come from hosting and the third-party messaging platforms it connects to.\n\nMaturity: high for a community/open-source project. It has versioned releases, migration guides, i18n/Crowdin integration, community resources, contribution guidelines, and a plugin publishing workflow. The presence of experimental features (e.g., Loader service, permissions, interaction, theme development, language packs, bundles) shows continued evolution but also some beta surface area.\n\nIntegrations: the main integration surface is the adapter layer to messaging platforms, plus database connectors and optional HTTP/server services. It also supports Satori, an open protocol for chat bot interoperability, which increases its reach beyond the built-in adapters."
}2/3 tests passed
| Test | Endpoint | Status | Latency |
|---|---|---|---|
| website_uptime | GET / | 200 | 163ms |
| robots_txt | GET /robots.txt | 404 | 36ms |
| llms_txt | GET /llms.txt | 200 | 41ms |
{
"overall": 72,
"dimensions": {
"token_efficiency": 7.0,
"first_try_success": 6.0,
"response_parseability": 7.0,
"error_clarity": 6.0,
"doc_quality": 7.0,
"auth_simplicity": 9.0,
"latency": 9.9,
"consistency": 7.0
},
"pricing_normalized": {
"model": "free",
"open_source": true,
"self_hosted_cost": "infrastructure-dependent",
"note": "No paid tiers; costs are infrastructure and third-party messaging platform requirements."
},
"issues": [
"No robots.txt (404) — may hinder crawlers",
"No security headers detected on responses",
"Documentation appears China-centric (Chinese locale default, adapters like DingTalk/QQ/WeCom)",
"Free/open-source with self-hosting means hidden infrastructure cost and ops burden",
"No llms.txt well-structured but depth of plugin docs varies"
],
"recommendations": [
"Add robots.txt and security headers for better crawler/agent compatibility",
"Provide clearer English-first quickstart and value-prop summary",
"Publish an explicit pricing/TCO guide for self-hosting",
"Improve structured error/limitation docs for plugin and adapter setup"
]
}Show your live agent-readiness score on your own site. Free, no auth — it updates as your score changes.
<a href="https://prowl.world/service/koishi">
<img src="https://prowl.world/badge/koishi.svg" height="56" alt="Agent-readiness on Prowl">
</a>
See operational metrics, LLM evaluations, agent readiness, and more.
Open in Dashboard