One MCP config entry. 150+ specialist agents on demand. No manual setup.
Your AI assistant is a generalist. Sometimes you need a specialist -- a game economy designer, a security auditor, a technical writer. This MCP server gives your assistant instant access to 150+ expert agent templates. Describe what you need, it finds the right agent and spawns it.
You: "Help me design a balanced game economy"
Claude: [searches -> finds Game Economy Designer -> spawns it -> expert response]Templates auto-fetch on first run from agency-agents and stay updated. You don't touch a thing.
You can. The agency-agents install script copies all 160+ agent files directly into your tool's config directory (e.g. ~/.claude/agents/). It works -- but every agent's name and description is loaded into the context window of every conversation, whether you use them or not.
We measured it:
Approach | Context cost | When |
Installed agents ( | ~8,300 tokens | Every conversation, always |
MCP server (idle) | ~55 tokens | Every conversation |
MCP server (searching) | ~350 tokens | Only when you search |
MCP server (using an agent) | ~2,700 tokens | Only when you spawn one (median) |
That's a 150x reduction in baseline context usage. You get the same 160+ agents, but you only pay for the one you're actually using.
Installed agents (8,300 tokens): We ran the agency-agents install script (install.sh --tool claude-code), which copied 162 agent files to ~/.claude/agents/. Then opened a fresh Claude Code session and ran /context. Claude Code reported "Custom agents: 8.3k tokens" -- loaded into every conversation regardless of whether any agent is used.
MCP idle (55 tokens): With the MCP server configured instead, /context shows only the two deferred tool names (agency_search, agency_browse) and a brief server description in the system prompt. No agent data is loaded.
MCP searching (350 tokens): Measured by tokenizing the full JSON tool schemas that get loaded when the assistant calls ToolSearch to resolve the agency_search and agency_browse tools. Counted with @anthropic-ai/tokenizer.
MCP using an agent (2,700 tokens): The median token count across all 145 agent files, measured with @anthropic-ai/tokenizer. Only the single agent file you're actually using gets loaded into context. The range is 383–12,724 tokens depending on the agent (p25: 1,549, p75: 3,584).
Related MCP server: pantheon-mcp
As a plugin:
/plugin marketplace add npupko/agency-mcp-server
/plugin install agency@agency-mcp-serverOr via CLI:
claude mcp add agency -- npx -y agency-mcp-serverAdd to your MCP config:
{
"mcpServers": {
"agency": {
"command": "npx",
"args": ["-y", "agency-mcp-server"]
}
}
}That's it. First launch clones templates to ~/.cache/agency-mcp-server/ and pulls updates every 24 hours.
Ask your assistant:
"Search for a game economy designer agent"
You should see results from the agency_search tool. If it's the first run, templates will auto-download (~30 seconds).
Your assistant gets four tools:
agency_search(query, division?) -- describe a task, get matching agents with spawn instructions
agency_browse(division?) -- explore divisions and agents when you want to see what's available
agency_status() -- check index freshness: agent count, last update time, whether an update is available
agency_update() -- pull latest templates from git and rebuild the search index without restarting
When you ask for help with something specific, your assistant calls agency_search, picks the best match, and spawns a subagent with that specialist's full system prompt. You get an expert response without ever touching a config file.
Agents are organized into divisions:
Division | Examples |
Engineering | Software Architect, DevOps Engineer, Technical Writer |
Design | UI Designer, UX Researcher, Design Systems |
Game Development | Game Economy Designer, Game Mechanics Designer |
Marketing | Content Strategist, SEO Specialist, Email Marketing |
Security & Specialized | Security Auditor, Data Scientist, Legal Analyst |
...and more | Academic, Sales, Strategy, Support, Testing, Spatial Computing |
All configuration is through environment variables in your MCP config:
Variable | Default | Description |
|
| Path to agent templates. Set this to use your own templates instead of auto-cloning |
|
| Git repo to clone templates from. Point at your fork |
|
| Set to |
|
| Hours between update checks |
Point at a local directory:
{
"mcpServers": {
"agency": {
"command": "npx",
"args": ["-y", "agency-mcp-server"],
"env": {
"AGENCY_AGENTS_PATH": "/path/to/your/agent-templates"
}
}
}
}Or clone from your own repo:
{
"mcpServers": {
"agency": {
"command": "npx",
"args": ["-y", "agency-mcp-server"],
"env": {
"AGENCY_REPO_URL": "https://github.com/yourorg/custom-agents.git"
}
}
}
}Each agent is a Markdown file with YAML front-matter, organized by division:
engineering/
software-architect.md
devops-engineer.md
design/
ui-designer.md
game-development/
game-economy-designer.md---
name: Software Architect
description: Expert software architect specializing in system design...
---
Full agent system prompt goes here.The server indexes the name and description fields for search. The full Markdown body becomes the agent's system prompt when spawned.
agency_search(query, division?) -- find agents by task description, returns matches with file paths and a ready-to-use spawn template
agency_browse(division?) -- list all divisions, or list agents within a specific division
agency_status() -- check index freshness: agent count, last update time, whether an update is due
agency_update() -- pull latest templates from git and rebuild the search index at runtime
agency://agents -- full agent index as JSON
agency://divisions -- division list with counts and examples
use-agent -- describe a task, get the best-matching agent with spawn instructions
npm install
npm run build
# Run with auto-fetched templates
node dist/index.js
# Run with local templates
AGENCY_AGENTS_PATH=./my-agents node dist/index.js
# Type checking
npm run typecheck
# MCP Inspector
npm run inspectAgent templates from agency-agents by @msitarzewski.
MIT
Browse all agent divisions and their agents. Explore the agent registry when you want to see what's available. Use agency_search instead if you already know what kind of agent you need. Call with no arguments to see all divisions. Pass a division name to list its agents.
| Name | Required | Description | Default |
|---|---|---|---|
| division | No | Division to list agents for (omit to see all divisions) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate read-only, non-destructive, idempotent behavior. The description adds clarity on how to invoke different behaviors (no args vs division), but does not add novel behavioral traits beyond annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise, well-structured with usecase and instructions tags, and front-loaded with the primary action.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given low complexity (1 optional param, no output schema), the description provides complete guidance on usage and alternatives, leaving no gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% with a clear description for the division parameter. The description restates the schema's intent without adding new semantic detail, meeting the baseline.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'Browse all agent divisions and their agents.' It differentiates from sibling agency_search by recommending its use when knowing the agent type.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicit instructions: 'Call with no arguments to see all divisions. Pass a division name to list its agents.' Also includes when to use agency_search instead.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Find and launch a specialized agent for any task. Search agent templates by keyword. Returns matching agents with file paths and a spawn template. Call this before spawning any agency subagent.
Pass a task description as query (e.g. 'game mechanics', 'security audit')
Pick the best match from results
Spawn a subagent using the template at the bottom — replace with the file path and <describe the user's task> with the user's full, unabridged request
Return the subagent's response directly to the user without summarizing it
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | Task or keyword to search for (e.g. 'game mechanics', 'frontend React', 'security audit') | |
| division | No | Optional: narrow to a division (e.g. 'engineering', 'game-development') |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate readOnlyHint=true, idempotentHint=true, destructiveHint=false. The description adds behavioral details like output format (matching agents with file paths and spawn template) and the spawning workflow. It does not contradict annotations and provides useful context beyond them.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is well-structured with <usecase> and <instructions> tags, front-loading the main purpose. Each sentence adds value, though the instructions are detailed. It is concise for the complexity involved.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Despite lacking an output schema, the description comprehensively explains the output (matching agents with file paths and spawn template) and provides full workflow instructions. Given the tool's complexity and the annotations covering safety, the description is complete enough for an AI agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% with both parameters described. The description adds example values for query (e.g., 'game mechanics') and division (e.g., 'engineering'), and clarifies that query should be a task description, enhancing the schema's meaning.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'Find and launch a specialized agent for any task' and the usecase elaborates on searching agent templates by keyword, returning file paths and spawn templates. It distinguishes from siblings (agency_browse, agency_status, agency_update) by focusing on search and spawning, not browsing, status, or updates.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly states 'Call this before spawning any agency subagent' and provides step-by-step instructions on how to use it: pass task description, pick best match, spawn using the template, and return response directly. This gives clear when-to-use and how-to-use guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Check the current status of the agent index — last update time, whether an update is available, and agent count.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare the tool as read-only, non-destructive, and idempotent. The description adds valuable behavioral details: what specific data the tool returns (last update time, update availability, agent count), which goes beyond the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, clear sentence with no fluff. It front-loads the purpose and efficiently conveys the key information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (0 params, no output schema), the description fully informs the agent of what the tool does and what to expect. It covers all necessary aspects for correct invocation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are no parameters, so the description does not need to add param meaning. The baseline for 0 params is 4, and the description effectively explains the output, compensating for the absence of an output schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb 'check' and the resource 'agent index status', and specifies the three pieces of information returned (last update time, update availability, agent count). This distinguishes it from sibling tools like agency_browse or agency_search.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies when to use (for a quick status check) but does not explicitly state alternatives or when not to use. No guidance on context or exclusions is provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Pull latest agent templates from git (if applicable) and rebuild the search index.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already provide idempotentHint=true, but description adds context: pulling from git (with 'if applicable') and rebuilding the search index. This clarifies the exact side effect beyond the annotation flags.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Single sentence, no fluff. Every word adds value: specifies action, resource, and condition ('if applicable'). Efficient and front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no parameters, no output schema, and a simple action, the description is sufficient. It covers the essential behavior and conditionality, making it complete for an agent to understand and invoke.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
No parameters in schema; schema coverage is 100%. Description adds no parameter info, but baseline for 0 parameters is 4. No need for additional parameter details.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Description clearly states the verb 'pull' and 'rebuild' on specific resources 'agent templates' and 'search index'. Distinguishes from sibling tools (browse, search, status) as an update operation.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No explicit when-to-use or when-not-to-use guidance. However, the idempotentHint annotation implies it can be called repeatedly without side effects, and siblings handle other tasks. Lacks explicit alternatives or exclusion criteria.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Each tool has a clearly distinct purpose: browse lists divisions/agents, search finds agents by keyword with spawn templates, status checks index health, update refreshes the index. No overlap.
All tools follow a consistent 'agency_' + verb in snake_case pattern (browse, search, status, update), making it predictable and easy to understand.
With 4 tools, the server is slightly on the minimal side but still well-scoped for agent registry operations. Each tool serves a distinct purpose without redundancy.
The tool surface covers the core workflows: browsing, searching, status checking, and updating. Minor gap is the lack of a direct spawn tool, but search provides a template for spawning.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
150+ vertical AI expert bots as agent tools. $1 bots run on YOUR machine - your data stays yours.
AI agent tools: web search, browser, 400+ LLMs, image gen, TTS, phone verify. Pay-per-use.
The cloud for agents. Tools for AI agents to register, build, and deploy other agents. Zero human required.
Build, validate, and deploy multi-agent AI solutions from any AI environment.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/npupko/agency-mcp-server'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
