MCP-native · Semantic tool routing · 120+ servers

Stop Wasting Tokens. Start Using the Right Tools.

Reduce token costs by up to 90% in minutes.

Like LangChain, but optimized for tool efficiency. Agent-CoreX dynamically discovers, ranks, and executes the best tools for your AI — reducing latency, cost, and noise.

No credit card required

Open source · MIT license

Works with OpenAI, Anthropic, MCP Servers

agent-corex · tool retrieval

89.4%

Tokens Saved

100 → 4

Tools Selected

63%

Faster

Live system benchmarks

Real-Time Performance Gains

Live metrics from real usage and system benchmarks

89.4%

~$0.28 / request

Tokens Saved

63%

avg across workflows

Latency Reduced

100 → 4

per avg request

Tool Calls Reduced

2,847,192

and counting

Requests Processed

96.2%

across all executions

Tool Success Rate

1200+

MCP tools indexed

Tools Available

100+

MCP servers registered

Top MCP Servers

githubfilesystembrave-search

3.8%

low-confidence tools filtered

Rejection Rate

How Agent-CoreX Works

Three steps to cut token waste and boost AI performance — automatically.

01

Retrieve Tools

Agent-CoreX receives your query and scans 120+ indexed MCP tools using semantic embeddings to surface the most relevant candidates instantly.

02

Rank & Filter

A cosine-similarity ranking pass scores every candidate. Only top-confidence tools pass the threshold — irrelevant noise is discarded before it ever touches your context.

03

Execute Efficiently

The filtered tool set (typically 3–5 tools) is injected into your AI context and executed via MCP JSON-RPC — delivering results with 63% lower latency and 89% fewer tokens.

Entire pipeline runs in 10–100ms with Redis cache · no extra infra required

From Chaos to Precision

The same AI request. Radically different outcomes.

Before
No tool routing — all 100 tools loaded
get_customer_data
send_email
process_payment
check_weather
get_recipes
send_sms
get_stock_price
order_pizza
+ 92 more tools …
Tools injected100
Avg latency850ms
Tokens / req28,400
Monthly cost$30K
After · Agent-CoreX
4 tools selected · ranked by semantic score
github_deploy_tool
0.97
docker_build_tool
0.94
k8s_rollout_tool
0.91
ci_status_tool
0.88
96 tools filtered out
Tools injected4
Avg latency315ms
Tokens / req2,840
Monthly cost$8K

$22K / month saved

73% cost reduction · 63% faster

Same query. Same AI. Completely different cost profile.

Live Activity

Real-time tool execution events across the platform

agent-corex · live events
LIVE
session totals:tokens saved: 284,719cost saved: $1284.52
00:00:01TOKEN SAVEReduced tokens by 82% on query "search GitHub issues"
00:00:03EXECExecuted github_list_prs_tool via MCP JSON-RPC
00:00:05COST SAVESaved $0.42 on single request — 96 tools filtered out
00:00:07FILTEREDReduced tools 56 → 6 for query "deploy to kubernetes"
00:00:09EXECExecuted brave_web_search · latency 48ms (cache hit)
00:00:11TOKEN SAVEToken savings: 28,400 → 2,840 (90% reduction)
00:00:13COST SAVESaved $1.18 on batch of 3 queries
00:00:15FILTEREDReduced tools 120 → 4 for "read filesystem and summarize"

Built with Developer Trust

Production-grade architecture. Security by default. Zero magic boxes.

Built for Engineers

Type-safe SDK · OpenAPI spec

MCP-Native

JSON-RPC 2.0 · stdio transport

Secure Sandbox Execution

Per-user workspace isolation

SHA-256 API Keys

Keys never stored in plaintext

Multi-Layer Cache

9 Redis layers · 95%+ cache hit

Open Source Core

agent-corex CLI on GitHub

REST API · Bearer auth · JSON response
# retrieve top tools for your query
GET /retrieve_tools?query=deploy+app&top_k=5
 
# response
{
"tools": [
{ "name": "github_deploy_tool",
"score": 0.97,
"server": "github-mcp" },
{ "name": "docker_build_tool",
"score": 0.94,
"server": "docker-mcp" }
],
"tokens_saved": "89.4%",
"latency_ms": 48
}
Content-Type: application/json200 OK · 48ms
View on GitHub· Full API docs in dashboard

Simple, Transparent Pricing

From free open source to enterprise scale. Build with your own MCP server packs.

Self-Hosted

Deploy on your own infrastructure — free, unlimited, forever.

View on GitHub →

Free

Explore Agent-CoreX

1,000 queries / mo📦 1 Pack⚙️ 3 servers / pack
$0/mo
1,000 API queries / month
1 custom pack
3 MCP servers / pack
Tool playground
MCP server browser
Community support
MOST POPULAR

Pro

Best for Developers

200,000 queries / mo📦 5 Packs⚙️ 8 servers / pack
$29/mo

₹2,499/mo

200,000 API queries / month
5 custom packs
8 MCP servers / pack
Usage analytics & query history
Semantic tool retrieval
Email support

Team

For Growing Teams

500,000 queries / mo📦 Unlimited Packs⚙️ Unlimited servers
$99/mo

₹8,499/mo

500,000 API queries / month
Unlimited custom packs
Unlimited MCP servers
Team sharing (coming soon)
Advanced analytics
Priority support

Enterprise

Large-Scale Deployments

Custom query limits📦 Custom limits⚙️ Custom limits
Custom
Custom query limits
Custom pack & server limits
Dedicated infrastructure
SLA guarantee
Dedicated support team
On-prem deployment option

Compare plans

FeatureFreeProTeamEnterprise
API Queries / Month1K200K500KCustom
Custom Packs15Custom
MCP Servers / Pack38Custom
Tool Playground
Usage Analytics
Query History
Team Sharing
Priority Support
SLA / Dedicated Infra

All paid plans include instant activation · Secure Razorpay checkout · Cancel anytime

Questions? We've Got Answers

Everything you need to know about Agent-CoreX implementation and ROI.

Average setup time is 2 hours. Most teams are in production within 1 day. We handle the integration — you focus on your agents.

Still have questions? Let's talk.

Our team can answer any question and give you a personalized cost estimate.

Schedule a Call with an Engineer
No setup required · Works in minutes

Let Your AI Use Tools the Right Way

Stop burning tokens on irrelevant tools. Agent-CoreX routes every query to precisely the tools it needs — automatically, in milliseconds.

Schedule a Call
✓ No credit card required✓ Free tier available✓ Works with any LLM provider✓ 1.2k+ developers trust Agent-CoreX