Applied AI for software development, presented as an operating surface.
This variant treats the landing page like the control room itself. Pearl Computing turns AI-assisted software work into an inspectable runtime with status panels, queue state, repo-native identity, and explicit operator commands. Rune is the concrete runtime behind that model.
system framing
The common pattern in AI software work is simple: a transient chat, hidden context, and weak continuity. Pearl is built around the opposite assumption. Useful agent workflows need visible control surfaces, persistent state, and operator intervention points that feel native to engineering work.
That is why this page looks more like a dashboard than a generic marketing site. The presentation is the message: Pearl is for teams that want to run AI workflows as systems, not merely visit them as sessions.
workflow delta
disposable chat workflow
pearl runtime workflow
runtime topology
command surface
The commands are intentionally direct. Operators initialize a runtime, inspect status, connect to the repo, ask questions, and adjust execution paths without pretending the system is magical.
rune init <name>
Provision the runtime home and start the services that keep the agent loop durable.
rune status <name>
Inspect agent health, deployment information, and runtime endpoints.
rune clone-url <name>
Expose the managed repo so the operator can audit and evolve the system in source.
rune ask <name>
Query the active runtime without abandoning the existing system context.
rune model set-path
Persist local model configuration so inference remains close to the runtime.
runtime modules
identity module
Runtime behavior comes from source-controlled files, making prompts auditable and editable like software.
state module
Work can pause, resume, and survive beyond one operator session without a hard reset.
queue module
Queues and task surfaces make work routing visible instead of burying it in implied context.
control module
Humans can approve, redirect, and reconfigure the system without breaking continuity.
execution module
Inference stays near the machine and repo when trust, cost, or latency require it.
deployment module
Runtime state and behavior can travel from workstation experiments to managed environments.
queue state
Focused repo-aware coding work with preserved context.
Repeatable tests, grading passes, and artifact review.
Operational assistants grounded in policy, state, and source control.
Interfaces for redirection, approval, and runtime observation.
closing signal
If the goal is to move from impressive demos to dependable engineering systems, Pearl and Rune are aimed at the missing layer: a runtime that can be inspected, directed, and evolved in the open.