60.75K
total weekly downloads
54 packages
capybara
Capybara — the calm, enterprise-grade open-source agent platform. Coming soon.
@capybara-hq/device
capybara-device: connects a laptop or server to a Capybara deployment as an environment
@capybara-hq/kernel
Capybara kernel: config, database, plugin host and HTTP API of the Capybara agent platform
@capybara-hq/plugin-anthropic
Anthropic model provider (model-provider@2) using the official SDK and the native Messages API
@capybara-hq/plugin-ark
Model provider (model-provider@2) for Volcengine Ark chat and embedding models
@capybara-hq/plugin-bailian
Model provider (model-provider@2) for Alibaba Cloud Model Studio (Bailian) chat and embedding models
@capybara-hq/plugin-client
Capybara plugin client SDK: the types, hooks and design re-exports a plugin's client entry is written against
@capybara-hq/plugin-connector-mcp
MCP connector (connector@1): tools and resources of remote MCP servers over Streamable HTTP
@capybara-hq/plugin-connector-openapi
OpenAPI connector (connector@1): one tool per operation of an OpenAPI 3.0 / 3.1 HTTP API
@capybara-hq/plugin-deepseek
Model provider (model-provider@2) for DeepSeek chat models through the Chat Completions API
@capybara-hq/plugin-file-preview
Previews code, archives, spreadsheets, documents, PDFs, SVG and HTML files in the conversation; entries inside archives are previewed regardless of per-previewer settings
@capybara-hq/plugin-github
GitHub connector templates: issues, pull requests, reviews, repositories and Actions through GitHub's remote MCP server, as the connected person (GitHub App / OAuth App or a personal access token)
@capybara-hq/plugin-jina
Web search and page reading (web-search@1, web-fetch@1) through Jina's search and reader APIs
@capybara-hq/plugin-lark
Lark (channel@1 bot over the Lark long connection with interactive cards, identity-source@1 web login, directory@1)
@capybara-hq/plugin-memory
Memory for agents: each agent remembers each member across conversations, with an index at run start, memory tools, hybrid search and a memory page
@capybara-hq/plugin-model-fake
Deterministic scripted model provider (model-provider@2) for runtime tests, e2e tests and keyless demos
@capybara-hq/plugin-object-store-s3
Object storage (object-store@1) on any S3-compatible service (AWS S3, MinIO, Ceph, OSS, TOS, ...) using the official @aws-sdk/client-s3
@capybara-hq/plugin-openai
OpenAI model provider (model-provider@2) using the official SDK and the native Responses API
@capybara-hq/plugin-openai-compatible
Model provider (model-provider@2) for Chat Completions-compatible services (vLLM, DeepSeek, Qwen, ...) using the official openai SDK
@capybara-hq/plugin-sandbox-docker
Conversation sandboxes (sandbox@1) as Docker containers (runc or gVisor runsc) behind a Docker socket proxy, with an egress allowlist proxy
@capybara-hq/plugin-sandbox-e2b
Conversation sandboxes (sandbox@1) as E2B sandboxes (E2B cloud or self-hosted), with pause / resume and E2B egress controls
@capybara-hq/plugin-sdk
Capybara plugin SDK: the types and pure functions plugins are written against
@capybara-hq/plugin-skills-core
Core skills every agent can use: skill-guide teaches an agent to write, register, test and share its own skills
@capybara-hq/plugin-slack
Slack (channel@1 bot over Socket Mode with Block Kit cards, identity-source@1 Sign in with Slack, directory@1)
@capybara-hq/plugin-telegram
Telegram (channel@1 bot over the Bot API: long polling or webhook, inline keyboard cards)