MCP Standard Library

A comprehensive collection of 461 MCP tools for AI agents — text processing, data transformation, system operations, and agentic capabilities via sampling.

Works with any MCP client: Claude Code, Claude Desktop, Cursor, VS Code Copilot, and more.

Quick Start

Claude Code / Claude Desktop

Add to your .mcp.json:

{
  "mcpServers": {
    "std": {
      "command": "deno",
      "args": ["run", "-A", "jsr:@casys/mcp-std/server"]
    }
  }
}

Load specific categories only

{
  "mcpServers": {
    "std": {
      "command": "deno",
      "args": ["run", "-A", "jsr:@casys/mcp-std/server", "--categories=text,json,math"]
    }
  }
}

---

Building Custom MCP Servers

NEW in v0.2.0: Use the ConcurrentMCPServer framework to build your own high-performance MCP servers.

MCP std now includes a production-ready server framework with built-in concurrency control, backpressure, and sampling support. Perfect for building custom MCP servers with your own tools.

Quick Example

import { ConcurrentMCPServer } from "jsr:@casys/mcp-server";

// Define your custom tools
const myTools = [
  {
    name: "my_tool",
    description: "My custom tool",
    inputSchema: {
      type: "object",
      properties: {
        input: { type: "string", description: "Input data" }
      },
      required: ["input"]
    }
  }
];

// Create handlers
const handlers = new Map([
  ["my_tool", async (args) => {
    return `Processed: ${args.input}`;
  }]
]);

// Create server with concurrency control
const server = new ConcurrentMCPServer({
  name: "my-server",
  version: "1.0.0",
  maxConcurrent: 5,                    // Max 5 concurrent requests
  backpressureStrategy: 'queue'        // Queue excess requests
});

server.registerTools(myTools, handlers);
await server.start();

Framework Features

• Wraps Official SDK: Built on @modelcontextprotocol/sdk for full protocol compatibility
• Concurrency Control: Limit concurrent requests (default: 10)
• Backpressure Strategies:
  • 'sleep' - Busy-wait when at capacity (default)
  • 'queue' - FIFO queue for pending requests
  • 'reject' - Fail fast when at capacity
• Sampling Support: Optional bidirectional LLM delegation
• Metrics: Monitor in-flight and queued requests
• Production-Ready: Used by mcp-std (461 tools)

Configuration Options

interface ConcurrentServerOptions {
  name: string;                       // Server name
  version: string;                    // Server version
  maxConcurrent?: number;             // Default: 10
  backpressureStrategy?: 'sleep' | 'queue' | 'reject';
  backpressureSleepMs?: number;       // Default: 10ms (for 'sleep' strategy)
  enableSampling?: boolean;           // Enable LLM sampling
  samplingClient?: SamplingClient;    // Custom sampling implementation
  logger?: (msg: string) => void;     // Custom logger
}

Advanced Usage

import {
  ConcurrentMCPServer,
  RequestQueue,
  SamplingBridge
} from "jsr:@casys/mcp-server";

// Custom queue configuration
const queue = new RequestQueue({
  maxConcurrent: 15,
  strategy: 'queue',
  sleepMs: 10
});

// Get metrics
const metrics = queue.getMetrics();
console.log(`In-flight: ${metrics.inFlight}, Queued: ${metrics.queued}`);

// Sampling for agentic tools
const server = new ConcurrentMCPServer({
  name: "agentic-server",
  version: "1.0.0",
  enableSampling: true,
  samplingClient: myCustomSamplingClient
});

const bridge = server.getSamplingBridge();
const result = await bridge.requestSampling({
  messages: [{ role: 'user', content: 'Analyze this data...' }]
});

---

Agent Tools (Sampling)

Agentic MCP tools that delegate work to LLMs via the sampling protocol.

The agent category provides 8 powerful tools that leverage MCP Sampling to call LLMs from within tool execution:

Tool	Description
agent_delegate	Delegate a subtask to another agent with context
agent_analyze	Deep analysis of code, data, or documents
agent_classify	Classify input into categories
agent_summarize	Generate summaries at various detail levels
agent_extract	Extract structured data from unstructured text
agent_validate	Validate content against rules/schemas
agent_transform	Transform data with natural language instructions
agent_generate	Generate content from specifications

Sampling Compatibility

Client	Sampling Support	Notes
PML	Server-side	Handles sampling automatically — works everywhere
VS Code Copilot	Native	Full sampling support
Claude Code	Not yet	Tracking issue #1785
Claude Desktop	Not yet	Coming soon

Tip: Use with PML for sampling support in any MCP client.

---

Capabilities (Tool Composition)

Combine multiple MCP tools into reusable, named capabilities.

PML (Procedural Memory Layer) extends mcp-std with capabilities — compound tools that chain multiple operations:

// Example: A capability that fetches, transforms, and validates data
const result = await mcp.pml.pml_execute({
  intent: "Fetch user data and validate schema",
  code: `
    const data = await mcp.std.http_fetch({ url: "https://api.example.com/users" });
    const parsed = await mcp.std.json_parse({ text: data });
    const valid = await mcp.std.schema_validate({ data: parsed, schema: userSchema });
    return valid;
  `
});

Key features:

• Learn from execution — Successful tool chains become reusable capabilities
• Semantic search — Find capabilities by intent, not just name
• Version control — Track capability evolution over time
• Multi-tenant — Isolated capabilities per user/org

Learn more: github.com/Casys-AI/casys-pml

---

Tool Categories

System (85 tools)

Category	Count	Description
docker	19	Container lifecycle, images, compose, logs
database	16	PostgreSQL, SQLite, MySQL, Redis CLI access
sysinfo	12	CPU, memory, disk, network, processes
pglite	7	Embedded PostgreSQL (in-process, no server)
process	5	Exec, spawn, kill, signal handling
packages	5	npm, pip, apt, brew, cargo
git	4	Status, diff, log, commit, branch
archive	4	tar, zip, gzip, compression
kubernetes	4	kubectl get/apply/delete/logs
ssh	3	Remote exec, scp, tunnels
media	3	ffmpeg, imagemagick wrappers
cloud	3	AWS, GCloud, systemd

Data Processing (212 tools)

Category	Count	Description
format	25	Number, bytes, yaml, toml, markdown, SQL formatting
string	21	Trim, pad, truncate, wrap, case conversion
crypto	20	Hash, UUID, base64, JWT, HMAC, TOTP, bcrypt
collections	20	Map, filter, sort, unique, group, chunk, flatten
algo	20	Binary search, quickselect, top-n, sorts
math	17	Eval, stats, round, unit conversion, roman numerals
validation	11	Email, URL, UUID, IP, phone, credit card
text	10	Split, join, regex, template, slugify
json	10	Parse, query, merge, flatten, pick, diff
encoding	10	NEW ROT13, Caesar, Morse, NATO, binary, hex, base32, punycode
transform	8	CSV, XML, YAML parsing and stringification
datetime	7	Now, format, diff, add, cron parsing
diff	7	Unified diff, patch, compare lines
compare	6	Levenshtein, similarity, fuzzy match, deep equal
security	6	NEW JWT generate/verify, password strength, checksums, CRC32
textanalysis	4	NEW Readability scores, sentiment analysis, word frequency

Utilities (134 tools)

Category	Count	Description
color	19	Hex/RGB/HSL conversion, palettes, blend, contrast
path	13	Join, dirname, basename, resolve, relative
geo	13	Distance, bearing, bbox, geocode, timezone
util	11	HTTP status codes, MIME types, user agents
qrcode	10	Generate/decode QR codes, barcodes, SVG output
state	10	In-memory KV store with TTL, persistence
devtools	10	NEW Semver parse/compare/bump, roman numerals, env parser, cron, regex
vfs	8	Virtual filesystem for sandboxed operations
resilience	8	Retry, rate limit, circuit breaker, timeout
network	8	URL parsing, IP info, subnet calc, DNS
iptools	7	NEW CIDR calculator, subnet divide, IPv6, MAC format
http	6	Build URLs, parse headers, query strings
schema	6	Infer JSON schema, validate, generate samples
timezone	5	NEW Timezone convert, world clock, meeting planner, DST

Generation (32 tools)

Category	Count	Description
faker	16	Mock data: names, addresses, companies, lorem
data	11	Generate images, SVG, placeholder data
python	5	Execute Python code, manage pip packages

Agentic (13 tools)

Category	Count	Description
agent	8	LLM delegation via sampling (see above)
pml	5	Capability management: list, lookup, rename, merge, whois

Note: The pml tools require a PML account. Set PML_API_KEY to authenticate.

---

TypeScript API

import { MiniToolsClient } from "jsr:@casys/mcp-std";

const client = new MiniToolsClient();

// List all tools
const tools = await client.listTools();
console.log(`${tools.length} tools available`);

// Call a tool
const result = await client.callTool("text_split", {
  text: "hello,world",
  separator: ","
});
// => ["hello", "world"]

// Get tools by category
const cryptoTools = client.getToolsByCategory("crypto");

---

Environment Variables

Variable	Description	Default
PGLITE_PATH	Path for embedded PGlite database	./data/pglite
MCP_STD_CATEGORIES	Comma-separated list of categories to load	all

---

Credits

This library includes tools inspired by:

• IT-Tools MCP
• TextToolkit MCP
• Math MCP
• JSON MCP

---

---

MCP Apps UI Components

Interactive UI components for MCP hosts using the MCP Apps extension (SEP-1865).

mcp-std includes 12 interactive UI components that display tool results visually. When an MCP host supports the Apps extension, these UIs are rendered alongside tool responses. Each UI is atomic and composable.

UI Stack

Layer	Technology	Purpose
Framework	Preact	Lightweight React-compatible (3KB)
Styling	Panda CSS	Build-time CSS-in-JS (0KB runtime)
Design System	Park UI preset	Tokens, colors, dark mode
Bundle	Vite + singlefile	Single HTML with inline CSS/JS (~105KB gzip)

Available UIs

Data Display

UI	Description	Associated Tools
table-viewer	Sortable, filterable table with pagination	psql_query, pglite_query, mysql_query, sqlite_query, docker_ps, docker_stats
json-viewer	Collapsible tree view with search and path copying	json_parse, json_query, schema_infer
chart-viewer	Bar, line, pie charts with interactivity	math_stats, collections_group

Code & Diff

UI	Description	Associated Tools
diff-viewer	Unified/split diff display with hunk navigation	git_diff, diff_lines, diff_unified
log-viewer	Filterable logs with level highlighting and auto-scroll	docker_logs

Metrics & Monitoring

UI	Description	Associated Tools
gauge	Circular/linear/compact gauge with thresholds	free (memory)
sparkline	Inline mini chart with trend indicator	uptime (load average)
metrics-panel	Grafana-style grid of metrics	df (disk usage)

Design & Input

UI	Description	Associated Tools
color-picker	Swatches, formats (HEX/RGB/HSL), palettes, WCAG contrast	color_parse, color_palette, color_contrast
form-viewer	Dynamic forms from JSON Schema with validation	schema_infer, validate_schema
qr-viewer	QR code display with download and copy	qr_generate_url
status-badge	Valid/invalid/warning badges with details	validate_email, validate_url, validate_ip

Tool → UI Mapping

Each tool declares its UI via _meta.ui.resourceUri:

// In tool definition
_meta: {
  ui: {
    resourceUri: "ui://mcp-std/gauge",
    emits: ["click"],      // Events the UI sends to model
    accepts: ["refresh"],  // Commands the UI accepts
  },
}

Complete mapping:

Category	Tool	UI
Database	psql_query	table-viewer
	pglite_query	table-viewer
	mysql_query	table-viewer
	sqlite_query	table-viewer
JSON	json_parse	json-viewer
	json_query	json-viewer
	json_compare	json-viewer
Math	math_stats	chart-viewer
Git	git_status	table-viewer
	git_log	table-viewer
	git_diff	diff-viewer
	git_branch	table-viewer
Diff	diff_lines	diff-viewer
	diff_unified	diff-viewer
	diff_words	diff-viewer
	diff_chars	diff-viewer
	diff_similarity	gauge
Compare	diff_text	diff-viewer
	diff_json	json-viewer
	diff_arrays	table-viewer
	compare_deep_equal	status-badge
Color	color_parse	color-picker
	color_palette	color-picker
	color_contrast	color-picker
System	df	metrics-panel
	du	table-viewer
	free	gauge
	uptime	sparkline
Docker	docker_ps	table-viewer
	docker_images	table-viewer
	docker_stats	table-viewer
	docker_logs	log-viewer
	docker_network_ls	table-viewer
	docker_volume_ls	table-viewer
Kubernetes	kubectl_get	table-viewer
	kubectl_apply	status-badge
	kubectl_logs	log-viewer
	kubectl_exec	log-viewer
Archive	tar_extract	table-viewer
	unzip	table-viewer
Schema	schema_infer	json-viewer
QR Code	qr_generate_url	qr-viewer
	qr_wifi	qr-viewer
	qr_vcard	qr-viewer
Validation	validate_email	status-badge
	validate_url	status-badge
	validate_ip	status-badge
HTTP	http_get	json-viewer
	http_post	json-viewer
	http_request	json-viewer
	http_parse_url	json-viewer
Network	curl_fetch	json-viewer
	ping_host	status-badge
	dig_lookup	table-viewer
	ip_address	table-viewer
	traceroute	table-viewer
IP Tools	cidr_calculate	json-viewer
	cidr_contains	status-badge
	subnet_divide	table-viewer
	ip_convert	json-viewer
	ipv6_expand	json-viewer
	mac_format	json-viewer
	ip_range	table-viewer
SSH	ssh_exec	log-viewer
	rsync	log-viewer
Packages	npm_run	log-viewer
	pip_run	log-viewer
	apt_install	status-badge
	apt_search	table-viewer
	brew_install	status-badge
Process	ps_list	table-viewer
Crypto	crypto_jwt_decode	json-viewer
Security	jwt_verify	status-badge
	password_strength	gauge
Text Analysis	text_readability	gauge
	text_sentiment_simple	gauge
	text_word_frequency	chart-viewer
	text_statistics	metrics-panel
Datetime	datetime_parse	json-viewer
	datetime_cron_parse	json-viewer
Timezone	tz_world_clock	table-viewer
	tz_meeting_planner	table-viewer
Geo	geo_nearest	table-viewer
	geo_distance_matrix	table-viewer
	geo_validate	status-badge
	geo_point_in_polygon	status-badge
Faker	faker_person	table-viewer
	faker_company	table-viewer
State	state_keys	table-viewer
	state_stats	metrics-panel
Collections	array_group	json-viewer
	array_count_by	chart-viewer

Event Communication

UIs emit events via structuredContent that can trigger PML sync rules:

// Table: user selects a row
{ event: "select", rowIndex: 5, row: { id: 123, name: "Alice" } }

// Log viewer: user filters by level
{ event: "filterLevel", levels: ["error", "warn"] }

// Gauge: user clicks metric
{ event: "click", id: "cpu", value: 73 }

// Color picker: user copies color
{ event: "copy", format: "HEX", value: "#ff5733" }

Building UIs

cd lib/std/src/ui
npm install
npm run build    # Builds all 12 UIs to dist/

Creating Custom UIs

1. Create folder: src/ui/my-viewer/
2. Add index.html and src/main.tsx
3. Import Panda CSS: import { css } from "../../styled-system/css"
4. Connect to MCP host:

   const app = new App({ name: "My Viewer", version: "1.0.0" });
   app.connect();
   app.ontoolresult = (result) => { /* handle data */ };

5. Build: npm run build

See src/ui/table-viewer/ for a complete example.

---

License

MIT