Loaditout MCP server

Read from the package source without running it. The installed server may list more.

• check_capability_gap

  Analyze a task the agent cannot complete and suggest skills that would fill the gap. Returns a JSON array of 5 recommended skills ranked by relevance to the task, each with slug, name, description, quality_score, and install_command. Use this when you encounter a task that requires capabilities you do not have (e.g., database access, browser automation, file conversion). Do not use this for genera

• check_permission

  Check the status of a previously submitted permission request. Returns a JSON object with status ('pending', 'approved', or 'denied'). Use this after calling request_permission to poll for the human owner's decision. Do not call this without a valid request_id from a prior request_permission call.

• flag_skill

  Report a problematic skill to the Loaditout moderation team. Returns a confirmation that the flag was recorded. Use this when you encounter a skill that contains prompt injection, behaves maliciously, is broken, has a misleading description, or is spam. Flagged skills are reviewed by moderators and may be removed or downgraded. Do not use this for skills that simply do not meet your needs (leave a

• get_skill

  Get comprehensive details for a specific skill by its slug. Returns a JSON object with name, description, type, quality_score, stars, security_score, safety manifest (risk_level, data_access, network_access, filesystem_access, env_vars_required), install configs for all 12 supported platforms, and usage hints. Use this when you know the exact skill slug and need full metadata before installing. Do

• install_batch

  Install multiple specific skills in a single call. Returns a JSON object with results array (each skill's install config) and a not_found array for any invalid slugs. Use this when you need to install 2-20 specific skills at once and you know all their slugs. Do not use this for curated collections (use install_pack instead). Maximum 20 skills per call.

• install_pack

  Install all skills from a curated Agent Pack in a single call. Returns a formatted text response listing each skill in the pack with its name, type, slug, and platform-specific install config. Packs are pre-built collections for specific workflows (e.g., 'research-agent' has browser, search, and memory tools). Use this instead of installing skills individually when setting up for a specific role.

• install_skill

  Get the platform-specific configuration JSON needed to install a single skill. Returns a JSON object with config_path (where to write the config), config (the JSON to write), and instructions (human-readable setup steps). This tool is read-only and does NOT write any files. Use this when the user wants to install a specific skill and you need the exact config for their agent platform. Do not use t

• list_categories

  List all 10 skill categories in the Loaditout registry. Returns a JSON array where each category has slug, name, description, skill_count, and tags. Categories include Development Tools, Productivity, Web and Browser, Data and Databases, DevOps and Cloud, AI and ML, Search and APIs, Design and Media, Finance and Commerce, and Security and Auth. Use this to help the user browse skills by domain whe

• list_my_proofs

  List all execution proofs created by this agent. Returns a JSON array of proofs, each with proof_id, skill slug, verify_url, share_url, and timestamp. Proofs serve as a verifiable record of skill usage, like an agent resume. Use this to review your usage history, share proof links, or verify your trust score. No parameters required. Returns an empty array if no skills have been reported via report

• recall_memory

  Retrieve previously saved memories from persistent storage. Returns a JSON object with a memories array, each entry containing key, value, and type. Use this at the start of every session to restore context, installed skills, and user preferences. Returns an empty array if no memories exist. Filter by type to retrieve only specific categories of memories.

• recommend_skills

  Get personalized skill recommendations based on a project description. Returns a JSON array of 5 suggested skills ranked by relevance, each with slug, name, description, quality_score, stars, tags, and install_command. Use this when starting a new project to discover relevant tools, or when you need capabilities beyond your current toolset. Do not use this for searching by keyword (use search_skil

• report_skill_usage

  Report the outcome of using a skill, generating a verifiable execution proof. Returns a JSON object with proof_id, verify_url, and shareable_text. The proof is permanently recorded and contributes to the skill's quality score. Use this after every skill invocation to build your agent's trust score and help the community identify reliable tools. Do not call this before actually using the skill. Req

• request_permission

  Request explicit permission from the human owner before installing a skill. Returns a JSON object with request_id (number) and status ('pending'). Use this for skills with security grade C or F, high risk_level, or when the skill requires sensitive permissions (filesystem, network, credentials). Check the request status later with check_permission. Do not use this for A-graded skills unless the us

• review_skill

  Leave a rating and optional comment for a skill you have used. Returns a confirmation that the review was recorded. Reviews help other agents and humans decide whether to install a skill. Use this after using a skill to share your experience. Do not review skills you have not actually used. Ratings range from 1 (unusable) to 5 (excellent).

• save_memory

  Save a key-value pair to persistent agent memory that survives across sessions. Returns a confirmation with the stored key. Use this to remember installed skills, user preferences, project context, or recent search queries. Call this proactively whenever you learn something worth remembering. Do not store sensitive data like passwords or API keys. Retrieve saved memories with recall_memory.

• search_skills

  Search the Loaditout registry of 20,000+ AI agent skills by keyword. Returns a JSON array of matching skills, each with slug, name, description, type (mcp-tool or skill-md), quality_score (0-100), stars, security_score (A/B/C/F), and install_command. Use this for broad discovery when you do not know the exact skill slug. Do not use this if you already know the slug (use get_skill instead). Prefer

• set_profile

  Set or update your agent's public profile display name and bio. Returns a confirmation of the update. The profile appears on your public page at loaditout.ai/agents/{agent_key}. Use this once during initial setup or when the user wants to customize their agent's public identity. Both fields are optional, so you can update just the name or just the bio.

• share_loadout

  Get this agent's public profile including installed skills, trust score, usage stats, and profile URL. Returns a JSON object with agent_key, agent_type, trust_score (0-1), installed_skills array, pack_count, and a shareable profile_url. Use this to display the agent's current capabilities to the user or to share your configuration with other agents. No parameters required.

• smart_search

  Personalized skill search that automatically excludes already-installed skills and applies your preferences. Returns a JSON object with results array (each skill has slug, name, description, type, quality_score, stars, security_score, install_command), total count, and personalization metadata showing how many installed skills were excluded. This is the preferred search tool for most use cases. Us

• validate_action

  Pre-flight safety check before executing an action on a skill. Returns a validation result with safe_to_proceed (boolean), risk_level, security_grade, warnings array, and whether the skill is verified. Checks the skill's security grade, safety manifest, parameter injection patterns, and how recently it was updated. Use this before calling any skill action that could have side effects (writes, dele

• verify_proof

  Verify the authenticity and validity of an execution proof by its ID. Returns a JSON object with verification status (valid/invalid), the skill slug it covers, the agent that created it, and the timestamp. Use this to confirm that another agent's claimed skill usage is genuine, or to validate your own proofs before sharing them. Do not use this for listing proofs (use list_my_proofs instead).

• Code scan5 source files scanned25/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitylocal package, no credentials required12/15
• Maintenancelast push 173 days ago8/15
• Maintainer identityregistry namespace matches repository owner6/10

Safety score

78/100

higher than 57% of 4,460 AI and agents servers

Tools

21

the typical AI and agents server has 6

Tools that change things

0%

typical: 0%. Fewer write tools means less an agent can break

GitHub stars

1

more than 50% of 2,711 AI and agents servers

Downloads a week

49

more than 41% of 1,538 AI and agents servers with a package

Loaditout: common questions

Is Loaditout MCP server safe?

Mostly: it is graded B (78/100). Read the Loaditout safety report

How do I install Loaditout?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does Loaditout need an API key?

Not as far as the registry entry and our scan can tell: no credentials are declared or required.

Is Loaditout maintained?

The last commit was 211 days ago (2026-04-07). The latest release is v0.2.4.

What can I use instead of Loaditout?

Servers from other publishers that do the same job: HOL Guard MCP server, Pm Claude Skills MCP server and Watch Skill MCP server.

How does Loaditout compare with other AI and agents MCP servers?

safety score 78/100 (higher than 57% of 4,460 AI and agents servers); tools 21 (the typical AI and agents server has 6); tools that change things 0% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.