A local-first desktop AI assistant for conversation, memory, voice, automation, and software work.
Website | Technical documentation | Issues | MIT License
Command bridge![]() |
Eva brings cloud and local models into one focused desktop experience while keeping configuration, approvals, and personal data under your control.
- Unified model routing: use GitHub Copilot ACP, OpenAI, OpenAI-compatible endpoints, or local models through LM Studio.
- SI execution: choose Eva Direct, Copilot CLI, or the experimental Pi runtime with approved tools for a configured OpenAI-compatible SI endpoint.
- Durable local memory: retain explicit facts with provenance, corrections, lifecycle controls, and an interactive memory view.
- Agent Observatory: inspect active agents, coding workspaces, mission history, security events, and memory from a visual command bridge.
- Tools and automation: run approved skills, research, scheduled work, terminal tasks, and MCP integrations with visible outcomes.
- Voice and messaging: use local speech features and optional Signal integration.
- Workspace control: create or import projects, monitor coding agents, and manage independent terminal sessions from the desktop app.
curl -fsSL https://appatalks.github.io/eva-agent/get-eva.sh | bash
evaEva is added to the system application menu. Select Eva (AIG) on first launch for the integrated routing, memory, and tools experience.
For source builds and platform requirements, see the technical documentation.
Current development version: 5.7.4 (unreleased)
Development build artifact: Eva Standalone-5.7.4.AppImage
Latest published standalone release: 5.6.10
Eva stores chat state, memory, and runtime data locally by default. Cloud providers, network tools, and autonomous coding features are used only when configured or enabled. Consequential actions remain subject to authorization and confirmation boundaries, and action receipts distinguish accepted work from verified completion.
Autonomous tools can access user-authorized workspaces or network resources. Review the runtime behavior and trust boundaries before enabling them.
- Technical documentation covers setup, providers, memory, voice, tools, architecture, and security boundaries.
- Standalone guide covers installation, packaging, updates, and runtime compatibility.
- Development guide covers project structure, contribution workflow, and validation.
- Testing contracts defines behavior, security, and compatibility expectations.








