Turn goals into loops.Keep what matters moving.
A self-hosted, local-first personal AI runtime connecting models, agents, durable state, workflows, automations, and every surface you use. You own the data, keys, and environment.
# One command — macOS, Linux, and Windows.
curl -fsSL https://xopc.ai/install.sh | bashWorks on macOS, Linux, and Windows. The one-liner installs Node.js and everything else for you. Hackable clones from git for local hacking. Switch tabs for npm or desktop downloads.
XOPC Agents
One runtime.
Three ways to deliver work.
Let agents keep working within the devices, data, identity, and workflows you own—then choose the entry point for the job at hand.
XOPC Worker
Hand over a goal. The agent plans, acts, asks at key decision points, and gives the result back to you.
Meet the computer agentFor software engineeringXOPC Code
Connect coding agents to a local repository and receive reviewable, verifiable changes.
Meet the coding agentAs a local gateway serviceXOPC Gateway
Connect a local xopc runtime to web, mobile, messaging channels, and APIs as a service.
Meet GatewayOne runtime.
Work that has a next step.
xopc separates long-running work into explicit state, execution, triggers, and access. It does not assume AI remembers everything; important information has a home, complex execution is observable, and follow-up can start again.
Durable state
Sessions, Projects, Goals, Notes, and Workspaces retain conversations, progress, source material, and artifacts.
Explicit execution
Agents use allowed tools and skills; Workflows organize multi-step or parallel work into inspectable runs.
Active triggers
Automations run agents or workflows on schedules, manually, or from webhooks, with visible success, failure, and retry history.
Access anywhere
Terminal, web, desktop, mobile, messengers, and HTTP/SSE APIs connect to the same runtime and state.
Chat is a surface.
The runtime is the product.
xopc puts models, agents, durable state, and execution into an environment you control instead of piling features into another chat window.
Runtime and ownership
Your hardware, keys, models, and deployment.
Your machine. Your rules.
xopc runs on your hardware. Conversations, project context, and local state stay in your xopc runtime. API keys stay in your config. No mandatory cloud. No surprise bills.
LOCAL FIRSTBring your own API keys. Choose any model.
DeepSeek, OpenAI, Anthropic, Google, Ollama, LM Studio, vLLM, Bedrock, Azure, and 20+ providers are built in. Run fully local inference or mix cloud and local models; switch with one config line and avoid vendor lock-in.
Durable state
Unified memory carries across agents while project context keeps clear ownership.
Unified memory, not one context window.
Sessions, projects, goals, notes, workspaces, and run history have explicit ownership. User memory, preferences, and connected knowledge accumulate in one place for reuse across agents and conversations, while project and workspace context keep their own boundaries.
Projects, Goals, and Notes have distinct jobs.
Projects group context, Goals track status, blockers, and next actions, while Notes capture text, voice, images, and attachments wherever input happens.
Execution and boundaries
Complex work is observable, repeated work is triggerable, and capabilities are configurable.
Observable workflows. Traceable automations.
Organize complex execution into phases and subagents, then trigger it on schedules, manually, or from webhooks. Progress, results, failures, and retries stay visible.
Agent capability boundaries are explicit.
Give each agent its own identity, responsibilities, model roles, workspace, tool and skill policies, workflows, and runtime boundaries. Rules for confirmation, forbidden actions, and escalation are explicit and reusable through shared settings; memory belongs to the unified user context.
Add capabilities only when needed.
Skills provide task methods, Extensions add tools, channels, and UI, and Connectors bring in external capabilities while the core stays clear.
Every surface.
One runtime.
Terminal, browser, desktop, mobile, and messengers share agents, sessions, and project state. The surface changes; your runtime and data ownership do not.
Full keyboard control, streaming output, zero perceived latency.
Bring your own keys.
Keep the system unlocked.
Cloud, local, or hybrid models. Your keys, your data, your deployment rhythm.
Complex execution,
visible at every step.
Workflows use explicit phases and parallel subagents for audits, research, and multi-perspective reviews; each run records progress, results, and failures.
style review · docs review