A/B Testing & Experimentation
How do I get started with Cline’s Plan/Act workflow on an existing repo without it making risky changes?
7 min read
Most teams approaching a plan/act AI workflow for the first time are worried about the same thing: “What happens if it touches my real code?” The safest way to get started on an existing repo is to treat Cline’s workflow like you would a new engineer on their first day—give it a clear sandbox, narrow permissions, and a review layer between its suggestions and your main branch.
Quick Answer: The Cline platform was acquired by Strictly AI, and the original plan/act workflow product is no longer available via cline.ai. To explore that workflow safely on an existing repo, you’ll need to work through the acquiring organization’s resources or any successor tools they provide, and always route changes through branches, PRs, and code review.
Frequently Asked Questions
What happened to Cline’s Plan/Act workflow and where do I start now?
Short Answer: The Cline platform was acquired by Strictly AI, and the cline.ai domain now functions as a transition notice rather than a product gateway. To get started with a plan/act-style workflow, you should refer to the official acquisition announcement and any successor documentation from Strictly AI.
Expanded Explanation:
If you arrive at cline.ai expecting to onboard Cline’s Plan/Act workflow on your existing repository, you’re hitting the domain after a change of ownership. The site now serves a single purpose: to confirm the acquisition (“The Cline platform was acquired by Strictly AI.”), link to an official announcement, and provide a contact for domain-related questions. It does not host product docs, onboarding flows, pricing, or technical guides.
Practically, that means there is no supported way to “sign up” or configure the original Cline Plan/Act workflow via this domain. Any current or future access paths, SDKs, or integrations would be controlled and documented by Strictly AI or in the canonical announcement it references. Until and unless those channels are published, the safest assumption is that the original workflow is not available for new implementations.
Key Takeaways:
- Cline’s Plan/Act product is not accessible or documented on cline.ai after the acquisition.
- Your next step is to review the official announcement and any resources it links for updated guidance.
How can I safely evaluate a plan/act-style AI workflow on an existing repo?
Short Answer: Use a sandboxed copy of your repo, restrict write access, and route all AI-generated changes through branches and code review before they reach your main line of development.
Expanded Explanation:
Even though cline.ai no longer provides a live Plan/Act workflow, the implementation pattern is familiar from other AI-assisted development tools. On an existing repository, you minimize risk by separating experimentation from production: work on a clone or fork, contain credentials, and never give direct write access to your protected branches.
The “plan/act” concept—first draft a plan, then execute in small, inspectable steps—maps well to standard Git practices: branches for experiments, pull requests for inspection, and status checks for automated validation. This keeps any AI-assisted changes reviewable and reversible, whether you’re using a successor to Cline or a different platform entirely.
Steps:
- Create a non-production copy of your repo (local clone, fork, or dedicated sandbox branch) and remove or mock sensitive secrets.
- Configure the AI tool to work only against that copy, ensuring it cannot push directly to your main or release branches.
- Enforce code review and tests by merging AI-generated changes via pull requests, with human approval and CI checks required before integration.
How does a Plan/Act workflow differ from letting an AI edit files directly?
Short Answer: A Plan/Act workflow separates “decide what to do” from “apply changes,” while direct file editing lets the AI modify code immediately without an explicit, reviewable plan.
Expanded Explanation:
In a Plan/Act pattern, the AI first proposes a structured change list—what files to touch, what behaviors to modify, and why—before any edits occur. Only after you review or approve that plan does it move into the “act” phase, applying incremental updates that you can track in version control. This mirrors how you’d onboard a new developer: first they describe their approach, then they implement in small steps.
By contrast, direct file editing compresses planning and execution into one step. The AI reads the repo and writes changes immediately, making it harder to understand intent, scope, or unintended side effects until after the fact. On an existing production repo, that lack of separation is precisely what most teams want to avoid.
Comparison Snapshot:
- Option A: Plan/Act workflow: Changes are preceded by an explicit plan, executed incrementally, and naturally routed through review.
- Option B: Direct editing: The AI modifies files on the fly, often without a human-approved checklist of expected changes.
- Best for: Teams working on existing codebases who need traceability, auditability, and a clear view of why each change exists.
How would I implement a Plan/Act-style process on my repo to avoid risky changes?
Short Answer: Mirror the Plan/Act idea using your existing Git flow: require a written change plan, keep AI actions confined to branches, and enforce human review with automated checks before merging.
Expanded Explanation:
Even without the original Cline workflow, you can reproduce its safety characteristics with the tools you already use. Treat “plan” as a required artifact (e.g., in an issue, PR description, or markdown doc) that lists the intended edits. Treat “act” as a controlled series of commits that implement that plan, each small enough to review. This minimizes surprises and aligns AI-driven changes with the same governance you use for human contributors.
The level of effort is modest if you already work with Git, branches, and CI. Most of the “implementation” is policy: no commits to main without a written plan and a PR; no secrets in the environment where the AI runs; and no bypassing tests just because the changes were AI-assisted.
What You Need:
- A Git workflow that uses branches, pull requests, and protected main/release branches.
- A policy that requires written plans, human review, and automated tests for all AI-generated changes.
Strategically, how should we think about Plan/Act workflows on a legacy codebase?
Short Answer: Treat Plan/Act as a risk-control layer: it lets you apply AI assistance to a legacy repo while preserving your existing safeguards around review, testing, and ownership.
Expanded Explanation:
Legacy repositories often have undocumented behaviors and fragile dependencies; this is exactly where unbounded AI edits are most dangerous. A Plan/Act-style workflow is valuable because it doesn’t replace your process—it fits into it. Plans become a form of design artifact, incremental acts become contained experiments, and your usual review and testing gates continue to protect production.
From a business standpoint, that means you can explore AI assistance without asking the organization to accept a “black box” editing your source directly. You keep your existing accountability structures in place—issue tracking, PR approvals, audit logs—while using AI to do scoped refactors, documentation updates, or repetitive edits under supervision.
Why It Matters:
- It reduces the risk of regressions in critical systems by keeping AI changes visible, reviewable, and testable.
- It allows you to experiment with AI-assisted development without compromising your existing governance or compliance posture.
Quick Recap
Cline’s original Plan/Act workflow is no longer accessible through cline.ai, which now serves only as an acquisition notice and routing page. If you want to apply a plan/act-style approach to an existing repo today, you’ll need to do so via tools and documentation provided by the acquiring organization or alternative platforms. In all cases, the safest pattern is the same: work in a sandbox or branch, require a written change plan, route all AI-generated edits through pull requests, and keep your existing review and testing gates in place before anything reaches production.
Next Step
Get Started](https://cline.ai)