Shynk Documentation

Sonnet-class quality at Haiku pricing. One API for domain-specific boosts, PII protection, and model routing.

What is Shynk?

Shynk is AI middleware that sits between your application and LLM providers (Anthropic, OpenAI). It adds:

  • Data Shield — automatic PII redaction (emails, SSNs, credit cards, phone numbers) before data reaches the LLM
  • Boosts — domain-specific system prompts that dramatically improve output quality for specialized tasks
  • Model Routing — when a boost is active and you request auto, Shynk runs it on Haiku instead of Sonnet whenever the benchmarks show Haiku reaches the same quality, so you pay less for the same result

Shynk works via MCP (Model Context Protocol) for AI agents or a REST API compatible with the OpenAI SDK.

Quick Start

Get running in under 2 minutes. Choose your integration method:

MCP (Recommended)
Python
Node.js
cURL

Claude Desktop / MCP Client config

{
  "mcpServers": {
    "shynk": {
      "url": "https://api.shynk.io/mcp/sse",
      "headers": {
        "Authorization": "Bearer sh_your_api_key"
      }
    }
  }
}

That's it. Your AI agent now has access to Shynk tools: boost application, PII redaction, and the full boost catalog.

Python (OpenAI SDK compatible)

import openai

client = openai.OpenAI(
    base_url="https://api.shynk.io/v1",
    api_key="sh_your_api_key"
)

response = client.chat.completions.create(
    model="claude-sonnet",
    messages=[{"role": "user", "content": "Review this contract..."}],
    extra_headers={"x-provider-api-key": "sk-ant-..."},
    extra_body={"shynk": {"boost": "contract-analyzer"}}
)

print(response.choices[0].message.content)

Node.js

const response = await fetch("https://api.shynk.io/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer sh_your_api_key",
    "x-provider-api-key": "sk-ant-...",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "claude-sonnet",
    messages: [{ role: "user", content: "Review this contract..." }],
    shynk: { boost: "contract-analyzer" }
  })
});

const data = await response.json();
console.log(data.choices[0].message.content);

cURL

curl https://api.shynk.io/v1/chat/completions \
  -H "Authorization: Bearer sh_your_api_key" \
  -H "x-provider-api-key: sk-ant-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet",
    "messages": [{"role": "user", "content": "Analyze this contract..."}],
    "shynk": {"boost": "contract-analyzer"}
  }'
Get your API key: Sign up to get your sh_ API key.

Authentication

All API requests require two keys:

HeaderValuePurpose
AuthorizationBearer sh_...Your Shynk API key (identifies your account, plan, usage)
x-provider-api-keysk-ant-... or sk-...Your LLM provider API key (Anthropic or OpenAI). Shynk never stores this.

Shynk acts as a transparent proxy — your provider key is forwarded directly and never logged or stored.

MCP Integration

Shynk implements the Model Context Protocol (MCP), the open standard for connecting AI agents to external tools.

Endpoints

EndpointMethodTransport
https://api.shynk.io/mcpPOSTJSON-RPC 2.0 (stateless)
https://api.shynk.io/mcp/sseGETServer-Sent Events (streaming)
https://api.shynk.io/mcp/healthGETHealth check

Claude Desktop Configuration

{
  "mcpServers": {
    "shynk": {
      "url": "https://api.shynk.io/mcp/sse",
      "headers": {
        "Authorization": "Bearer sh_your_api_key"
      }
    }
  }
}

Generic MCP Client

// Initialize handshake
POST https://api.shynk.io/mcp
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "method": "initialize",
  "params": { "capabilities": {} },
  "id": 1
}

// Response:
{
  "jsonrpc": "2.0",
  "result": {
    "protocolVersion": "2024-11-05",
    "serverInfo": { "name": "shynk", "version": "0.2.0" },
    "capabilities": { "tools": {}, "resources": {} }
  },
  "id": 1
}

MCP Tools

shynk.boost.list

Returns all available boosts with their name, category, and tier.

// Request
{ "jsonrpc": "2.0", "method": "tools/call",
  "params": { "name": "shynk.boost.list", "arguments": {} }, "id": 2 }

// Response includes array of boosts:
[
  { "id": "contract-analyzer", "name": "Contract Analyzer", "category": "Legal", "tier": "premium" },
  { "id": "data-privacy-analyzer", "name": "Data Privacy Analyzer", "category": "Compliance", "tier": "premium" },
  ...
]

shynk.boost.apply

Applies a domain-specific boost to your prompt. The boost injects an expert system prompt.

// Request
{ "jsonrpc": "2.0", "method": "tools/call",
  "params": {
    "name": "shynk.boost.apply",
    "arguments": {
      "boost_id": "contract-analyzer",
      "prompt": "Review this software license agreement: ..."
    }
  }, "id": 3 }

shynk.compress

Applies Data Shield — redacts PII (emails, SSNs, IBANs, credit cards, phone numbers) and reports token savings.

// Request
{ "jsonrpc": "2.0", "method": "tools/call",
  "params": {
    "name": "shynk.compress",
    "arguments": {
      "prompt": "Contact [email protected] or call +1-555-0123 for the contract signed by SSN 123-45-6789"
    }
  }, "id": 4 }

// Response:
{
  "compressed": "Contact [EMAIL_REDACTED] or call [PHONE_REDACTED] for the contract signed by SSN [SSN_REDACTED]",
  "pii_redacted": 3,
  "tokens_saved": 8,
  "actions": [
    { "type": "EMAIL", "field": "content" },
    { "type": "PHONE_US", "field": "content" },
    { "type": "SSN", "field": "content" }
  ]
}

shynk.health

Returns service status and version info.

MCP Resources

shynk://boost-catalog

Full boost catalog as a JSON resource. Useful for agents that want to browse available boosts.

{ "jsonrpc": "2.0", "method": "resources/read",
  "params": { "uri": "shynk://boost-catalog" }, "id": 5 }

REST API — Chat Completions

OpenAI-compatible endpoint. Drop-in replacement for any OpenAI SDK usage.

POST /v1/chat/completions

FieldTypeRequiredDescription
modelstringYesModel ID: claude-sonnet, claude-opus, gpt-4o, etc.
messagesarrayYesOpenAI-format messages array
streambooleanNoEnable SSE streaming
shynk.booststringNoBoost ID to apply (e.g. "contract-analyzer")

Response

Standard OpenAI-format response with additional shynk metadata:

{
  "id": "chatcmpl-...",
  "choices": [{ "message": { "role": "assistant", "content": "..." } }],
  "usage": { "prompt_tokens": 150, "completion_tokens": 450 },
  "shynk": {
    "data_shield_actions": [{ "type": "EMAIL", "field": "content" }],
    "boost_applied": "contract-analyzer",
    "tokens_saved": 12
  }
}

REST API — Boosts

GET /v1/boosts

List all active boosts. Requires authentication.

curl https://api.shynk.io/v1/boosts \
  -H "Authorization: Bearer sh_your_api_key"

// Response:
{
  "boosts": [
    { "id": "contract-analyzer", "name": "Contract Analyzer", "category": "Legal", "tier": "premium", "version": "8.0" },
    ...
  ]
}
Note: Boost system prompts are proprietary and never exposed through the API. You see the name, category, tier, and version — the boost content is applied server-side.

Data Shield

Data Shield automatically runs on every request. It detects and redacts PII before your data reaches the LLM provider:

TypePatternExample
EMAILEmail addresses[email protected][EMAIL_REDACTED]
SSNSocial Security Numbers123-45-6789[SSN_REDACTED]
IBANInternational Bank Account NumbersDE89370400440532013000[IBAN_REDACTED]
CCCredit Card Numbers4111-1111-1111-1111[CC_REDACTED]
PHONE_USUS Phone Numbers+1-555-0123[PHONE_REDACTED]
PHONE_INTLInternational Phone Numbers+44 20 7946 0958[PHONE_REDACTED]

Redacted values are replaced by short placeholders, so a redacted prompt is usually slightly shorter than the original. The saving depends entirely on how much PII the document contains.

Boost Catalog

Loading the live catalog…

This table is loaded live from the Shynk API every time you open this page, so it always matches what is actually available to buy.

How to Use Boosts

Via MCP

// 1. List available boosts
{ "method": "tools/call", "params": { "name": "shynk.boost.list" } }

// 2. Apply a boost to your prompt
{ "method": "tools/call", "params": {
    "name": "shynk.boost.apply",
    "arguments": { "boost_id": "contract-analyzer", "prompt": "Review this contract..." }
}}

Via REST API

// Include boost in your chat completion request
{
  "model": "claude-sonnet",
  "messages": [{ "role": "user", "content": "Review this contract..." }],
  "shynk": { "boost": "contract-analyzer" }
}

Before & After Example

Without boost (generic response)

User: "Review this NDA between Company A and Company B"
Assistant: "This NDA appears to be a standard mutual non-disclosure agreement..."

With contract-analyzer boost

User: "Review this NDA between Company A and Company B"
Assistant:
## NDA Review Summary
**Type:** Mutual NDA | **Duration:** 3 years + 2-year survival
**Risk Level:** MODERATE

### Key Findings
1. **Scope (Section 2):** Overly broad definition of "Confidential Information"
   includes publicly available data — recommend narrowing.
2. **Non-compete (Section 7):** 24-month non-compete is likely unenforceable
   in California — flag for legal review.
3. **Remedies (Section 9):** One-sided injunctive relief favoring Company A...
...
Boost system prompts are Shynk's trade secret. You see the result, not the prompt. This ensures consistent quality and prevents prompt leaking.

Plans

Shynk plans give you access to the API, core features, and included boosts. All plans include Data Shield and basic analytics.

Feature Free Starter $29/mo Pro $69/mo Business $199/mo
Requests/month10K250K1M5M
Data Shield
Boost Composer
AnalyticsBasicBasicFullFull
Industry Insights
Audit Logs
SupportCommunityCommunityEmailPriority

Enterprise — custom pricing for unlimited requests, SSO/SAML, HIPAA, SOC2, on-premise, dedicated CSM, 99.9% SLA. Contact sales.

Boost Pricing

Boosts are priced individually based on quality, reliability, and domain value. Each boost has a monthly subscription price.

Boost Tiers

Tier Price Range Typical Use
Basic$3–8/moProductivity, communication, general-purpose enhancements
Standard$8–19/moTechnical, marketing, sales, customer service, DevOps
Premium$19–49/moLegal, finance, compliance, security — high-stakes domains
Elite$49–99/moComing soon — healthcare, legal compliance, aerospace

Quality Metrics on Cards

Each boost card shows two quality indicators:

  • Quality Score (e.g. 9.75) — median score across multiple test runs by Vilda (QA). Scale 0–10.
  • Variance Badge — how consistent the boost performs:
    • ≤0.05 Highly Stable — near-identical output quality every time
    • ≤0.15 Stable — reliably good with minor variation
    • ≤0.30 Good — solid performance, some variation expected

How Pricing Works

Boost subscriptions are separate from your Shynk plan. Your plan covers API access and core features; boosts are add-ons you subscribe to individually.

  • Browse the Boost Marketplace to see all available boosts with prices
  • Subscribe to any boost — it activates immediately on your API key
  • Use a boost by adding "shynk": {"boost": "boost-id"} to your chat completion request
  • Cancel anytime — no lock-in
Every boost is quality-verified. Before appearing in the marketplace, each boost passes at least 6 independent test runs scored by our QA system. Only boosts with median score ≥9.2 and variance ≤0.30 are published.

Ready to get started?

Start free — 10K requests/month, no credit card required.

Get Started
Be Shynk. — shynk.io