EXPLYT / FIELD NOTES
Practical stories about code quality and speed, product updates, and engineering perspectives from the team building an IDE-native AI agent.
Explyt Team9 Sept 20268 min readExplyt 5.19 opens subscription minutes to Claude Code, Cursor, Codex, Cline, OpenCode, and Kilo Code through an OpenAI-compatible gateway, fixes the context indicator, and renews MCP server tokens automatically.
Developers paste logs into the chat while the agent runs only git. Why that relay filters the evidence, and what direct access needs.
Yurii Kostyukov13 min readExplyt 5.19 opens subscription minutes to Claude Code, Cursor, Codex, Cline, OpenCode, and Kilo Code through an OpenAI-compatible gateway, fixes the context indicator, and renews MCP server tokens automatically.
Explyt Team8 min readOne Kafka symptom, many defects, one invariant about async TID processing. What our agent did, what it missed, and why the debugger matters.
Yurii Kostyukov9 min readAI coding agents in JetBrains IDEs report commits, files and green builds that never happened. Documented cases, the mechanisms behind them, and the one-minute checks that catch each one.
Yurii Kostyukov14 min readSame bug, same model, same prompt. The Superpowers systematic-debugging skill burned 132k tokens and produced a patch I discarded; Explyt's Debug skill used 67k and shipped.
Explyt Team7 min readA practical framework for deciding when to delegate software tasks to an AI agent using certainty, verifiability, error cost, and pilot evidence.
Explyt Team28 min readA practical evidence checklist for reviewing cross-module refactoring with JetBrains symbol resolution, IDE refactorings, run configurations, debugger state, and MCP logs.
Explyt Team8 min readConnect Claude Code and other MCP clients to JetBrains IDE navigation, refactorings, run configurations, and debugging with Explyt MCP.
Explyt Team7 min readExplyt 5.17 adds IDE-powered Move and Delete refactoring, a free MCP server for external agents, and chat branching.
Explyt Team10 min readWhat impressed me most is how well the agent understands the existing project. Its plans fit naturally into what is already built, even when I am working outside my usual tech stack.
Senior Analyst Developer · DXC Technology
LSP gives an agent coordinates; the full IDE gives it the project model. That native IDE context makes all the difference on real Java code.
Software Architect & AI Board Lead · Siemens
IDE-native stepping makes the difference on stateful Java applications.
Software Engineer
Explyt 5.19: your subscription in Claude Code, Cursor, Codex, and other tools
Frequently Asked Questions
