Agents as a Service
Universal Discovery LayerArchitecture Preview

One Search.Every Agent Asset.

The Agentic Index is a single MCP server that indexes the entire agentic ecosystem — skills, agents, MCPs, APIs, tools, models, datasets, and more. Your agent asks one question. The index returns the best asset for the job.

Scale

The Entire Ecosystem. Indexed.

0+
Total Assets
0
Asset Types
0
Data Sources
0
MCP Tools
Problem

The Discovery Bottleneck

Today, agentic assets are scattered across disconnected registries. Your AI agent mid-task cannot query across all sources to find the optimal tool. It must know which registry to search, use the right tool, parse the right format.

aaas-vault5,400+ skills
aaas-blocks12 code blocks
aaas.blog800+ entities
npm / GitHub1,000+ MCP servers
Local harness4,300+ components

The Agentic Index eliminates this fragmentation.

Taxonomy

17 Asset Types

Every type of agentic resource under one unified schema. Quality-scored. Relationship-aware.

skill5,441
aaas-vault
agent956
harness + blog
tool800
aaas.blog
mcp-server1,200
npm + github
model340
aaas.blog
code-block12
aaas-blocks
plugin275
harness
template15
harness
api-endpoint500
registries
script200
aaas.blog
integration180
aaas.blog
dataset150
aaas.blog
benchmark90
aaas.blog
paper120
aaas.blog
provider85
aaas.blog
hardware45
aaas.blog
execution-file100
harness
MCP Surface

11 Agent-Native Tools

Discovery

search

Hybrid keyword + semantic search across all 17 asset types. Faceted filtering by type, category, platform, quality tier.

resolve

"I need to send emails from my agent" — returns ranked assets with relevance scores and reasoning.

lookup

Get full details for any asset by ID or slug. Every field, every relationship, every score.

browse

Paginated navigation by type, category, or domain. Like a file browser for the entire ecosystem.

Intelligence

recommend

Given your current stack + a task description, suggests what you're missing.

stack

Recommend a complete agent + tools + skills + MCPs stack for a use case. Full solution in one call.

compare

Side-by-side comparison of 2-5 assets. Scores, capabilities, compatibility, trade-offs.

related

Relationship graph for any asset — what it requires, what it enhances, what alternatives exist.

System

stats

Index health: total assets, type distribution, quality tiers, freshness, source coverage.

compatibility

Check if assets work together — platform, version, dependency, and runtime validation.

submit

Submit a new asset for review. Authenticated agents can enrich the index.

Intent Resolution

Ask What You Need. Get What Works.

Agent Request
resolve("I need auth for my Next.js app with Google OAuth")
Index Response (ranked)
#1
code-block:auth/clerk95%

Version-verified Next.js 16 block, one-command install

#2
skill:clerk-nextjs-auth91%

5,400+ word skill guide with examples and patterns

#3
code-block:auth/authjs88%

Auth.js v5 with Google provider, JWT sessions

#4
mcp-server:auth072%

Enterprise OAuth, higher setup complexity

The agent picks auth/clerk, routes to aaas-blocks MCP for installation. Discovery + installation in one flow.

Sources

5 Aggregated Data Sources

aaas-vault

5,400+ production-ready AI agent skills. 52 categories. Quality-scored Gold/Silver/Bronze.

skill5,441
aaas.blog

800+ entities across 11 types. Composite-scored leaderboard. 15 self-healing agents.

tool, model, agent, benchmark, dataset, paper, provider, hardware, integration800+
Claude Code Harness

4,300+ local components — core agents, pool agents, skills, plugins, templates.

agent, skill, plugin, template4,326
aaas-blocks

12 composable, version-verified code blocks for Next.js 16. Three-layer architecture.

code-block12
MCP Registries

npm @modelcontextprotocol/* packages + awesome-mcp-servers + community registries.

mcp-server1,200+

Data source ready   Ingestion building

Search

Hybrid Search Pipeline

Not just keywords. Not just vectors. Both — fused with Reciprocal Rank Fusion, boosted by quality tier.

1
Keyword

FTS5 full-text search with BM25 scoring and Porter stemming

2
Semantic

1536-dim vector similarity via sqlite-vec — understands intent, not just words

3
Merge

Reciprocal Rank Fusion combines both rankers into a single score

4
Filter

Apply type, category, platform, and quality tier constraints

5
Boost

Gold-tier assets get 1.2x boost. Quality rises to the top.

6
Return

Top N results with facet counts for further refinement

Offline-firstEntire index runs locally in SQLite (~50MB). No network required. Falls back to aaas.blog API for live data.
Roadmap

Implementation Phases

1

Vault + Keyword Search

5,441 skills indexed with FTS5. search, lookup, browse tools live.

2

Multi-Source Aggregation

All 5 data sources ingested. Deduplication. 10,000+ entities.

3

Semantic Search + AI

Vector embeddings. Hybrid search. resolve, recommend, stack tools.

4

ask.blog Integration

Bidirectional sync. Unified search contract across MCP + web.

5

npm + Production

@supraforge/mcp-agentic-index on npm. Docker image. Pre-built index.

Pricing

Open Core. Premium Intelligence.

Open Source
$0

The full index — locally hosted, self-built. All 17 asset types, all 5 data sources.

  • +search, lookup, browse tools
  • +Keyword search (FTS5)
  • +Self-hosted SQLite (~50MB)
  • +Vault + harness + blocks ingestion
  • +npm install + rebuild yourself
  • +Community support
Premium
$29/mo

Hosted index with semantic search, real-time sync, and AI-powered intent resolution.

  • +Everything in Open Source
  • +resolve + recommend + stack tools
  • +Semantic vector search (1536-dim)
  • +Hybrid search with RRF fusion
  • +Real-time Firestore sync
  • +Pre-built index (zero rebuild time)
  • +MCP Registry ingestion (1,200+ servers)
  • +Priority support

Enterprise pricing for dedicated index instances, custom data sources, and SLA. Book a call

Ready to deploy?

Initialize Your
Autonomous Workforce

Start with the API — 500 free credits on signup. Or book a call to scope custom agent workflows for your business.