Versions:

0.19.55
0.19.53
0.19.52
0.19.37
0.19.32
0.19.31
0.19.26
0.19.23
0.19.16
0.19.5
0.19.1
0.18.2
0.18.1
0.17.59
0.17.55
0.17.46
0.17.42
0.17.26
0.17.16
0.17.15
0.17.1
0.17.0
0.16.10
0.16.9
0.16.7
0.16.5
0.16.3
0.16.1
0.15.11
0.15.7
0.15.6
0.15.3
0.15.0
0.14.23
0.14.22
0.14.21
0.14.20
0.14.17
0.14.10
0.14.9
0.14.6
0.14.5
0.14.3
0.14.0
0.13.28
0.13.24
0.13.16
0.13.15
0.13.4
0.13.0
0.12.26
0.12.23
0.12.1
0.12.0
0.11.1
0.11.0
0.10.29
0.10.27
0.10.21
0.10.19
0.10.10
0.10.2
0.9.40

Proma is an open-source, local-first AI desktop application published by ErlichLiu, currently at version 0.19.55 across a catalog of 63 recorded releases. It falls into the AI assistant and agent tooling category and is designed to bring a general-purpose agent experience into everyday professional workflows, targeting advanced users who need more than a simple chatbot. Built as a complete open-source implementation on top of the Anthropic Claude Agent SDK (@anthropic-ai/claude-agent-sdk), it combines multi-model chat, a general-purpose agent, workspaces, Skills, MCP server integration, remote bots, and memory capabilities in a single client. In Chat mode, users get multi-model conversations, attachment parsing, image input, Markdown, Mermaid, KaTeX and code highlighting, side-by-side conversations, system prompts, and context management. Agent mode supports workspace isolation, permission modes, file operations, streaming output for long-running tasks, plan confirmation, and follow-up questions, while complex jobs can be decomposed into sub-agents and tasks whose execution is shown in the message stream. Each workspace can independently configure Skills, MCP servers, and files, making it suitable for accumulating reusable capabilities. Proma natively supports Feishu and Lark bot bridging, with entry points provided for DingTalk and WeChat, so users can trigger local agent workflows from a phone or group chat, and it can flexibly connect to any large language model provider. Additional features include shared memory across Chat and Agent, web search, built-in chat tools, and agent recommendations. True to its local-first design, sessions, workspaces, attachments, configurations, and Skills are stored by default in ~/.proma/ as JSON and JSONL files without relying on a local database. The desktop experience is rounded out with auto-updates, proxy settings, file preview, global shortcuts, a quick task window, voice input, and light, dark, or system-following themes.

Tags:

人工智能 48

大语言模型 49

智能体 34

自主智能 30