Add ejentum-mcp under AI Alignment

[ejentum]

Adds ejentum-mcp under AI Alignment, alphabetically after the existing Circuit Breakers entry.

Why I think it fits the section. Circuit Breakers is a runtime steering technique that intervenes in model behavior to prevent unsafe outputs. The ejentum-mcp anti-deception harness is the same architectural shape (runtime intervention) targeting a different failure mode: sycophancy, validation pressure, authority-appeal compliance, and hallucination drift instead of jailbreaks. Both are pre-generation interventions, not post-hoc filters. The section currently has only one entry; ours expands the runtime-steering tooling cluster.

What it is. MCP server (npm ejentum-mcp, MIT, free tier 100 calls) exposing four cognitive harness tools (harness_reasoning, harness_code, harness_anti_deception, harness_memory). Each call returns a structured cognitive scaffold the model ingests before generation: a named failure pattern, an executable procedure, suppression vectors that block the obvious shortcut, and a falsification test for self-verification. The anti-deception harness is the directly alignment-relevant component; the entry frames around that even though the project includes three other modes.

Format. Matched the existing Tools section style ([Name](url) \Lang` - description (Website)`) — I noticed the section uses both terse and elaborate entries depending on the project's surface; opted for elaborate since AI Alignment is a thin section and a one-line description helps readers decide if it fits their use.

Repo: MIT licensed, single repo containing the MCP server plus four SKILL.md files in /skills/, editor adapters in /editors/ (Cursor/Windsurf/Cline), and an Anthropic Claude Code plugin manifest. Install: stdio via npx -y ejentum-mcp, or hosted HTTPS at https://api.ejentum.com/mcp (Bearer auth via EJENTUM_API_KEY, free tier 100 calls).

Maintainer is me; happy to address review feedback.

[ejentum]

Refreshed the entry in the same branch (commit just pushed) with:

1. Dual-language tags (Node.js Python TypeScript) reflecting twelve native framework integrations published on PyPI / npm since this PR was opened (CrewAI, Agno, PydanticAI, smolagents on PyPI; ejentum-ai, ejentum-mastra, ejentum-langgraph, ejentum-genkit on npm; LangChain Python, LlamaIndex, Letta, AutoGen scheduled).

2. Cleaner description: replaced the original "free tier 100 calls" phrasing with project description language only. Free and paid tiers documented on the website rather than the listing.

Happy to relocate to the Safety section instead of AI Alignment if that's a better fit (the harness_anti_deception mode and the listed entries in Safety like NeMo Guardrails, Garak, wildguard align cleanly). Let me know which you prefer.