The latest articles on DEV Community by LikelyYou (@likelyyoumybae). https://dev.to/likelyyoumybae https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3801100%2F8cc4a4d9-fed2-4954-8fd2-6a69433251d8.png https://dev.to/likelyyoumybae en LikelyYou Sun, 05 Jul 2026 10:09:19 +0000 https://dev.to/likelyyoumybae/i-built-a-local-first-canvas-lms-agent-workflow-for-codex-and-claude-code-users-5fkl https://dev.to/likelyyoumybae/i-built-a-local-first-canvas-lms-agent-workflow-for-codex-and-claude-code-users-5fkl <p><a href="https://canvas-pilot.likelyou.com/articles/workflow-layer-above-canvas-mcp" rel="noopener noreferrer"></a></p> <p>I have been building Canvas Pilot, a local-first Canvas LMS AI agent for recurring coursework workflows.</p> <p>The main idea is not "another Canvas API wrapper." Canvas MCP servers and Canvas API clients can help an agent see Canvas. Canvas Pilot is the workflow layer above that:</p> <ul> <li>scan Canvas assignments into an approval plan</li> <li>stop before execution</li> <li>let the student approve selected work</li> <li>route approved items into recurring course workflows</li> <li>produce review-ready drafts, result files, and REPORT.md</li> <li>keep private course data and drafts local</li> </ul> <h2> Why workflow memory matters </h2> <p>Many courses repeat the same assignment shape every week.</p> <p>One course might always put the real spec on an external site. Another might always require the same reading annotation format. Another might use the same problem-source and PDF delivery workflow.</p> <p>If an agent has to rediscover that every week, you are still doing the orchestration. Canvas Pilot tries to make the repeated pattern durable.</p> <h2> The workflow </h2> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>scan Canvas -&gt; approval plan -&gt; student approval -&gt; approved workflow -&gt; review-ready output -&gt; REPORT.md </code></pre> </div> <p>The scan step does not start doing assignments. It writes a plan and stops. Execution only happens after the user approves selected items.</p> <h2> Who it is for </h2> <p>This public preview is for AI power users first.</p> <p>If you already know how to use Codex, Claude Code, or similar local agents, the product can remove a lot of repeated Canvas coordination. It can make recurring coursework feel close to a one-command workflow: scan, approve, execute, review.</p> <p>If you do not already know how to operate local agent workflows, this version will feel hard. I would rather be honest about that than pretend it is a polished no-code SaaS.</p> <h2> Local-first boundary </h2> <p>Credentials, cookies, real course identifiers, runs, private overlays, assignment inputs, and drafts stay on the user's machine.</p> <p>The public repo contains the generic framework and public-safe examples.</p> <p>GitHub: <a href="https://github.com/X-isdoingreat/canvas-pilot" rel="noopener noreferrer">https://github.com/X-isdoingreat/canvas-pilot</a></p> <p>Website: <a href="https://canvas-pilot.likelyou.com/" rel="noopener noreferrer">https://canvas-pilot.likelyou.com/</a></p> ai programming productivity opensource