Give your coding agent production-grade engineering patterns
exactly when it needs them.
AgentPedia delivers focused architecture, UI motion, database, and security guidelines directly to your AI agent through MCP. Stop stuffing thousands of lines of instructions into static `.cursorrules` files.
Reduction in Base Prompt Tokens
P99 Retrieval Latency (L1 Cache)
Single-Turn Snippet Relevance
Core Security & UI Blueprints
Why current AI coding workflows are broken
Models default to median GitHub training data. Trying to fix this with massive static rule files creates an unmaintainable context bloat tax.
1. The Craft & Quality Gap
LLMs generate code that functions, but lacks polish. They default to stiff linear CSS curves, write unindexed PostgreSQL queries, and frequently omit Row-Level Security, leaving endpoints vulnerable to IDOR attacks.
2. The Context Bloat Tax
Stuffing thousands of lines of instructions into a .cursorrules file degrades model attention ("needle in a haystack"), increases inference costs, and severely impacts the Time-to-First-Token generation speed.
3. The Synchronization Debt
As frameworks evolve (e.g., Tailwind v3 to v4, Next.js Pages to App Router), your static rule files drift out of date. Updating them requires manually editing prompt files across every single repository in your organization.
Enterprise-Grade Context Delivery
Engineered to preserve your model's context window while securely delivering high-quality blueprints.
Zero Context Bloat
AgentPedia's dynamic FastMCP footprint uses less than 120 tokens, preserving 96% of your model's reasoning window compared to static prompt files.
Offline L1 Cache
Zero network latency. AgentPedia maintains a sub-millisecond local SQLite cache of core blueprints so your agent never waits.
Strict Privacy
Your codebase never leaves your machine. AgentPedia operates strictly as a read-only knowledge provider and never ingests proprietary code.
AST-Audited Blueprints
Every pattern in our registry is rigorously analyzed against OWASP Top 10 vulnerabilities, ensuring your agent writes secure code by default.
Ready in < 30 seconds
We built AgentPedia to integrate flawlessly into your workflow. A single terminal command scaffolds the exact `.cursor/mcp.json` triggers your agent needs.
- 1
Run the NPX CLI
Automatically detects your stack (Next.js, FastAPI, Prisma) and configures the bridge.
- 2
Auto-Generated AGENTS.md
Injects a lightweight 20-line trigger file so your agent knows exactly when to call the MCP.
Empowering every discipline
Staff Frontend Eng.
Cares about motion physics, layout morphs, and accessibility.
- Enforces Emil Kowalski spring tokens
- Prevents layout shifts during rendering
Systems Architect
Focuses on correctness, tenant isolation, and event systems.
- Mandates PostgreSQL RLS policies
- Applies transactional outbox patterns
Full-Stack Builder
Moves fast building production apps with maximum efficiency.
- Gets senior-level code without prompt tuning
- Zero configuration workspace setup
The Four Knowledge Pillars
Real, production-tested blueprints delivered dynamically to your AI agent in 20-40 line chunks.
Emil Kowalski Drawer Modal Springs
Calibrated physics tokens eliminating stiff CSS transitions with physical mass, damping, and stiffness parameters.
mass: 0.8, stiffness: 260, damping: 24
PostgreSQL Multi-Tenant Row-Level Security
Enforces airtight database-level isolation so LLM queries can never leak cross-organization tenant records.
CREATE POLICY tenant_isolation ON records
USING (tenant_id = current_setting('app.tenant_id', true));
Transactional Outbox & Idempotency Keys
Prevents dual-write state corruption between SQL databases and message brokers (Kafka/RabbitMQ/Redis).
await db.$transaction([createRecord, insertOutboxEvent]);
Keep-a-Changelog Git Diff Synthesizer
Automatically formats Git diffs into structured release notes categorized by breaking changes, features, and fixes.
### 🛡️ Security - fix(rls): Parameterize tenant isolation
The AgentPedia Advantage
See what happens when an AI coding agent is equipped with curated, just-in-time knowledge.
Rigid Snap Transitions
Generic AI prompts generate mechanical linear curves with no mass deceleration, jarring users during navigation.
Natural Kowalski Fluid Springs
The agent calls search_guidelines() and automatically applies precise mass and stiffness parameters.
Why AgentPedia wins
| Feature | AgentPedia | Static .cursorrules | Linters (Sonar) |
|---|---|---|---|
| System Prompt Tax | < 120 Tokens | 2,000–5,000 Tokens | Zero |
| Execution Timing | Active & Just-in-Time | Passive / Static | Reactive (Post-code) |
| Central Updates | Instant Cloud Sync | Manual per-repo edit | Centralized policies |
| Frontend Craft Focus | Native spring tokens | Rare / Inconsistent | None |
Simple, transparent access
During our Alpha phase, all features are completely free. Lock in your spot.
Indie Builder
Perfect for solo developers looking to improve their frontend craft.
- Frontend Craft Blueprints
- IDE Integrations
- Standard L1 Cache
Pro Engineer
Full access to backend security, architecture, and infinite context.
- Everything in Indie
- Backend & Database Blueprints
- AST Security Auditing
- Infinite L1 Cache
Enterprise
For engineering teams maintaining private design systems.
- Private Custom Blueprints
- Custom AST Linting Rules
- SAML SSO
Curated for modern production stacks