Research data, MCP tools, and agent skills
Access the published articles as JSON or Markdown, search them through the read-only MCP server, or download engineering skill templates.
Connect an AI assistantDownload articles as JSON or Markdown
Every published article, including the diagrams as ordered text, source links, and editorial notes. No account or API key required.
Individual articles
- The agentic code factory: from request to releasable changeMarkdown
- Code review after agents: verify the change, not the explanationMarkdown
- Context engineering: repository knowledge, MCP, and skillsMarkdown
- Evaluating agentic engineering: quality, cost, and deliveryMarkdown
curl https://vibehaus.team/api/researchSearch and read articles through MCP
Connect an AI assistant to search the library, read an article, or retrieve a skill. The Model Context Protocol (MCP) gives assistants a standard way to call these tools. This server serves only the public content published here.
Streamable HTTP · stateless · no authentication
Server URL
https://vibehaus.team/mcpAdd this URL as a remote Streamable HTTP server in a compatible MCP client. Use your client’s documented server setup; this page does not install anything.
search_research- Find articles by keywords. An empty query lists all published articles.
read_article- Read a full article by its exact slug, including sources and workflow text.
get_skill- Retrieve a skill’s Markdown by slug. Reading it does not install or execute it.
Try asking your agent
“Search Vibe Haus research for code review. Read the article, then explain which checks would help our current review process. Cite the source.”
Clients use POST at the server URL. Browser GET requests return 405 because this endpoint does not provide an event stream. For ordinary browser access, use the JSON feed. The endpoint has a 32 KiB request limit and serves a fixed public catalog.
Download agent skill templates
Three self-contained SKILL.md templates. Read one first, adapt it to your repository, then install it using your agent’s documented skill setup.
Each download belongs in a folder matching its slug, with the filename SKILL.md. A skill adds a procedure; it does not grant access or authorize releases.
Engineering task contract
Turn a feature request into a clear task: outcome, constraints, acceptance examples, and checks.
engineering-task-contract/SKILL.mdEvidence-first code review
Review the actual diff, test a case that could reveal a bug, and report findings another engineer can act on.
evidence-first-code-review/SKILL.mdAgent workflow evaluation
Plan a fair comparison of an agent workflow, including quality, human effort, failures, and total cost.
agent-workflow-evaluation/SKILL.mdSources, permissions, and limitations
Original Vibe Haus synthesis and proposed engineering workflows, informed by the linked primary sources. These articles are not peer-reviewed studies or measured customer results. Examples and thresholds are illustrative unless explicitly attributed. Read our editorial approach.
You may use and adapt Vibe Haus’s original workflows and skill templates with attribution. Linked third-party material remains subject to its own terms. We link to AI Engineer’s original talks and other primary sources; we do not redistribute their transcripts.
The feed and MCP tools serve the same authored article catalog as the website. They do not access customer data, modify repositories, execute code, or connect to private systems. Public access helps agents retrieve and cite the material; it does not guarantee inclusion in any chatbot’s answers.
Research updated October 10, 2026. Skills version: October 10, 2026. Inspect downloaded instructions before use. Client compatibility and the broader tooling ecosystem can change.
Back to the research