Cline's agentic loopThe cycle an agent runs in: decide, call a tool, read the result, decide again — repeating until the goal is met or a stop condition fires.Full definition → is now a standalone service with a gRPC API, so you can script it, run many instances in parallel, or embed it in other tools instead of only using the VS Code extension.
Key takeaways · AI-distilled
Cline Core runs as a plain Node process with a gRPC API. A user interface connects as a gRPC client and subscribes to state changes and LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition → messages, while Core calls a host provider layer, for example to ask VS Code for linter output.
IDE versions of Cline can delegate a task to the CLI as an experimental multi-agentUsing several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.Full definition → with a fresh context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →. In this first version, Cline runs one sub-cline at a time and waits for it to finish.
Several frontends can attach to one running task, even over a network, so the author says a mobile app could remote-control a desktop Cline. At launch this worked only between JetBrains IDEs and the CLI, with VS Code waiting on a migration.
Cline for JetBrains was already built on Cline Core, and the CLI ships a generic host provider so embedders need not implement IDE integration themselves. Full gRPC documentation was still pending when the post was published.
Terms in this piece · Glossary
agentic loop — The cycle an agent runs in: decide, call a tool, read the result, decide again — repeating until the goal is met or a stop condition fires.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
