EveryDev.ai
Subscribe
Home
Developers

5,086+ AI companies

  • Radar
  • Trending
AI Tools by Topic
  • AI Coding Assistants
  • Agent Frameworks
  • MCP Servers
  • AI Prompt Tools
  • Vibe Coding Tools
  • AI Design Tools
  • AI Database Tools
  • AI Website Builders
  • AI Testing Tools
  • LLM Evaluations
Follow Us
  • X / Twitter
  • LinkedIn
  • Reddit
  • Discord
  • Threads
  • Bluesky
  • Mastodon
  • YouTube
  • GitHub
  • Instagram
Get Started
  • Users
  • Rate Tools
  • About
  • Editorial Standards
  • Corrections & Disclosures
  • Community Guidelines
  • Advertise
  • Contact Us
  • Newsletter
  • Submit a Tool
  • Start a Discussion
  • Write A Blog
  • Share A Build
  • Terms of Service
  • Privacy Policy
Explore with AI
  • ChatGPT
  • Gemini
  • Claude
  • Grok
  • Perplexity
Agent Experience
  • llms.txt
Theme
With AI, Everyone is a Dev. EveryDev.ai ©
    1. Home
    2. Developers
    3. vibesenpai

    vibesenpai

    vibesenpai is an open-source Claude Code plugin that reads Claude Code sessions stored locally and identifies where an agent went wrong, showing the user's own words and the relevant turns plus a fix that can be applied. It emphasizes evidence rather than scoring the developer or agent.

    Visit Website

    At a Glance

    1Tool Listed
    4Products
    9Capabilities
    Discussions
    2026Est.
    Focus Areas
    Vibe Coding
    Observability Platforms
    LLM Evaluations
    Connect
    Latest News
    Published a do-not-push demo video across the home page, README, and documentation; documentation site also added Umami page-visit measurement.Nov 17, 2026
    Released vibesenpai 0.1.0, adding consent-gated model-read previews, recorded/fingerprinted calls, the Status tab, and the Sent tab.Nov 16, 2026
    Markets
    • Claude Code users
    • Software developers using AI coding agents
    • Teams seeking local-first AI-agent observability
    • Developers who want evidence-based prompt and workflow corrections

    AI Tools by vibesenpai

    (1)
    View vibesenpai
    vibesenpai tool icon

    vibesenpai

    Claude Code Session Analyzer

    ObservabilityLLM EvaluationsVibe Coding

    Discussions

    No discussions yet

    Be the first to start a discussion about vibesenpai

    Latest News

    11/17/2026

    Published a do-not-push demo video across the home page, README, and documentation; documentation site also added Umami page-visit measurement.

    github.com
    11/16/2026

    Released vibesenpai 0.1.0, adding consent-gated model-read previews, recorded/fingerprinted calls, the Status tab, and the Sent tab.

    github.com
    11/14/2026

    Released 0.0.5 with a Rules tab and persistent records of findings dropped by the lookback gate.

    github.com
    11/14/2026

    Promoted 0.0.4 from alpha to stable version labeling.

    github.com

    Products & Services

    4
    vibesenpai Claude Code plugin
    2026-11-08 (0.0.1-alpha); 0.1.0 released 2026-11-16

    An MIT-licensed TypeScript Claude Code plugin installed through the vibesenpai marketplace. It reads local Claude Code session transcripts, runs free structural checks and optional consent-gated model reads, presents findings in an insights pane, and drafts fixes into the Claude Code prompt box rather than writing files itself.

    /vibesenpai:insights
    2026-11-08

    The side-pane command showing findings, the evidence turns, fixes, rules, status, and sent-call information.

    /vibesenpai:judge
    2026-11-08

    The command for previewing, enabling, or disabling optional model reads; model reads run through the user's own Claude account or API key.

    /vibesenpai:report
    2026-11-08

    Creates a self-contained offline HTML report containing findings, supporting turns, and fixes.

    Market Position

    vibesenpai positions itself as a local-first, evidence-oriented Claude Code plugin for post-hoc analysis of agent behavior: it shows the exact user instruction and surrounding turns, rather than assigning a score, and turns findings into an actionable fix. The ClaudePluginHub comparison page places it alongside broader Claude Code toolkits, multi-agent consultation plugins, and workflow/loop plugins; vibesenpai is narrower and specifically focused on diagnosing failures in prior sessions.

    Leadership

    Founders

    SR

    Souvik Roy

    Creator and primary maintainer of vibesenpai; GitHub identifies the repository's principal author as Souvik Roy (account crozzfire). The repository history shows him authoring the project and its releases; the project's launch description says he created it from frustrations with AI-driven development.

    Executive Team

    SR

    Souvik Roy

    Creator and primary maintainer

    Principal author of the public GitHub repository and author of the release and documentation commits.

    Founding Story

    The project was created after its author became frustrated with AI-driven development and wanted a tool that could show concrete evidence of agent failures in existing Claude Code transcripts and suggest a corrective prompt, rule, permission, guard, or test. It began under the name agentsensei and was renamed vibesenpai in the first alpha release.

    Business Model

    Revenue Model

    The plugin itself is open source and its free checks cost nothing. Optional model reads use Claude Haiku through the user's own Claude Code subscription or API key, so the documented cost is usage-based on the user's account rather than a separate vibesenpai subscription.

    Pricing Tiers

    Free checks
    $0

    Local structural checks run without charge.

    Optional model reads
    Usage-based; usually a few cents per session at most

    Runs on the user's own Claude Haiku plan or API key, within about 2 million tokens per day; the documentation gives about $0.26 as an API-price example.

    Target Markets

    Industries & Segments
    • Claude Code users
    • Software developers using AI coding agents
    • Teams seeking local-first AI-agent observability
    • Developers who want evidence-based prompt and workflow corrections
    Use Cases
    • Reviewing Claude Code sessions after an agent violated a stated constraint.
    • Finding misread requests, overreach, inaccurate claims, and repeated or failed tool calls.
    • Turning recurring agent mistakes into project-specific prompt rules, permissions, guards, or tests.
    • Monitoring and auditing local AI-assisted software-development workflows without uploading transcripts by default.
    • Inspecting history through an offline report.
    Notable Customers
    • Public demo based on test data from claude-code-log

    Quick Facts

    Founded
    2026

    History & Milestones

    2026-11-17

    The project published a recorded do-not-push demo video on the home page, README, and docs, and added Umami page-visit measurement to the documentation site.

    2026-11-16

    Version 0.1.0 introduced consent-gated model reads with an exact preview, recording/fingerprinting of model calls, a Status tab, and a paginated Sent tab.

    2026-11-14

    Version 0.0.4 dropped the alpha label; 0.0.5 added a Rules tab and retained records of lookback-gated findings.

    2026-11-10

    Versions 0.0.3-alpha and 0.0.4-alpha added documentation links in the pane and raised the minimum supported Claude Code version to 2.1.299.

    2026-11-09

    Version 0.0.2-alpha added support for Claude Code's CLAUDE_CONFIG_DIR so transcripts and settings in that configured directory are found.

    Key Capabilities

    9
    Reads Claude Code transcripts in place on the user's machine; transcripts are not copied by the plugin.
    Free checks for edit retries, repeated/redundant calls, and cache rebuilds.
    Optional model reads for constraint violations, misread intent, inaccurate self-report, overreach, and other problems.
    Model reads are off until the user sees an exact preview and explicitly turns them on.
    Findings lead with the developer's own words and cite the surrounding conversation turns.
    Evidence-based fixes can become permission prompts, guards, rules, tests, or read-only subagent requests.

    Integrations & Partnerships

    Platform Integrations

    • Claude Code plugin marketplace (/plugin marketplace add vibesenpai/vibesenpai).
    • Requires Node.js 22.13 or later and Claude Code 2.1.299 or later.
    • Uses the user's Claude Haiku access through a Claude subscription, API key, Bedrock, or Vertex when optional model reads are enabled.
    • MIT-licensed TypeScript source distributed through GitHub release zips; no npm installation or runtime dependencies are required.

    Key Partnerships

    The public demo credits test data from claude-code-log (MIT); this is a data/example-source relationship rather than a stated commercial partnership.

    Connect

    Website
    vibesenpai.github.io/vibesenpai/
    GitHub
    vibesenpai

    AI Topics

    3

    vibesenpai focuses on these topics:

    Vibe Coding(1)
    Observability Platforms(1)
    LLM Evaluations(1)
    Back to all developersSuggest an edit