The first Solution on Context Mode Platform — built on the open-source MCP plugin 400,000+ developers already run locally. One opt-in toggle turns local AI coding sessions into engineering signal your CTO, EM, CISO and FinOps read on Monday morning: per-engineer productive rate, retry waste, blockers named, role-narrowed views. Never source code, never prompts. 222 detection patterns, served as chat through MCP. self-serve · MCP-served.

Hi everyone — Mert here, maker of Context Mode Insight. Four months ago (first commit Feb 23, 2026) I started context-mode, an open-source MCP plugin that sandboxes tool output and saves AI coding agents ~98% of their context window. It grew to 400,000+ developers across 17 platforms (Claude Code, Cursor, Copilot, Codex, Gemini CLI, Aider…), hit #1 on Hacker News, and sits at 18,100+ GitHub stars. All organic. But the same question kept coming back from engineering leaders: "We run Claude Code, Cursor and Copilot across 50 engineers. My CTO asks on Friday what the $50K spend returned this quarter — and I don't have an honest answer." Insight is the opt-in org layer on top of the same plugin. Flip one toggle and what leaves the machine is structural events only — tool names, file paths, error and retry counts, session durations, commit outcomes. Never source code, prompts, or file content. Insight runs 222 patterns over those events and turns them into actionable answers — ask from the agent your team already runs, or read them straight off the dashboard: CTO › "What did our $50K AI coding spend return this quarter?" EM › "Who's blocked on the payments team this week — and on what?" EM › "Anyone trending toward burnout — late nights, weekend pushes, climbing retries?" EM › "Who shipped the most this sprint without review time ballooning?" IC › "How am I doing versus the rest of my team this sprint?" CISO › "Did any session touch secret-shaped paths in the last 24h?" FinOps › "Break our AI spend down by team for this month." DevOps › "Is commit quality holding, or are we shipping more reverts?" SecOps › "Flag the riskiest sessions across the org right now." Seven role-narrowed views (CTO, EM, IC, CISO, FinOps, DevOps, SecOps) — every answer an actionable insight (who to unblock, where spend is leaking, which risk to chase), with the full UI dashboard underneath for drill-down. On pricing: the plugin already saves your agents ~98% of their context per session — real money never spent. Insight reinvests a fraction of that back into the lens. The engine pays for the view. I'll be in the thread all day. The one thing I'd genuinely love to hear: which role view would you add for your org? Links: · Site: https://context-mode.com · Context Saving: https://context-mode.com/context... · OSS (ELv2): https://github.com/mksglu/contex... · Insight: https://context-mode.com/insight — Mert

Context is one of those things people don't notice until it's missing. It often makes the difference between an AI response that's helpful and one that completely misses the point.

This feels like exactly the layer engineering teams need right now. The strongest part for me is measuring real engineering impact without touching source code or prompts. Turning AI coding usage into honest ROI, bottleneck, and team insights is extremely valuable. The organic growth already proves how real the need is. Congrats, this is genuinely impressive.

Context is one of those things people don't notice until it's missing. It often makes the difference between an AI response that's helpful and one that completely misses the point.

Context Mode Insight

Engineering signal for AI-assisted teams

Engineering signal for AI-assisted teams