Gaia MCP server provides read-only tools for discovering and inspecting skills from the public Gaia Registry, plus server status monitoring. Use gaia_search to find generic and named skills by query, with filters for kind, tier, star rating, trust magnitude, contributors, installability, and types; returns ranked results with freshness metadata. Use gaia_inspect to get a detailed evidence-backed dossier for a specific skill, including relationships, implementations, trust signals, sources, and data freshness. Use gaia_status to check server version, registry mode, data freshness, counts, and available tools. (Note: The summon tool to materialize named skills into ephemeral sessions exists in the package but is not exposed in the current schema.)
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Gaia MCP Find a high-trust skill for automating email replies"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
DEPRECATED. This package is superseded by the Skill Heaven Claude Code
plugin, which bundles the summon MCP server directly — nothing to install
separately. See
gaia-research/gaia-skill-heaven.
The engine that used to live in this repo now lives at packages/skill-summon
in that monorepo, and ships esbuild-bundled straight into the plugin
(plugins/skill-heaven/mcp/skill-summon.mjs) — no npx, no sibling checkout,
no external binary, no runtime dependencies.
The published packages stay installable. @gaia-research/mcp and
skill-hell are not being unpublished — existing install.sh copies and
version-pinned installs keep working. Deprecated means don't start new work
here, not it broke. New work should target the plugin instead.
See DEPRECATION.md for the full closeout notes, including
the pending npm deprecate steps.
Agent-native Gaia Registry discovery, evidence inspection, and temporary skill summoning for Gaia Skill Tree.
Status:
@gaia-research/mcp@0.4.0is published. It is a working prototype: its current package interface is usable, while later Heaven/Summon profile work remains separate.
Gaia MCP consumes public Registry data; it does not own or mutate the Registry.
The published rich Registry/Bond package surface has four tools:
Tool | Current purpose |
| Find generic and Named Skills by task and constraints. |
| Return an evidence-backed skill dossier. |
| Materialize matching Named Skills into an ephemeral Skill Hell session. |
| Report version, Registry freshness, compatibility, counts, and available tools. |
summon is the current tool name; gaia_summon is not a current package tool.
It materializes skills in a temporary session and may maintain a bounded,
temporary cross-session payload cache. It never changes the Registry, current
repository, or permanent harness/project configuration.
The package's four tools are not an implementation or measurement of D4's
thin, two-tool Heaven/Summon profile (search_skills, summon). The latter is
a profile and schema-dose constraint for Skill Heaven work. It does not rename,
remove, or deprecate any of the published package tools above.
Related MCP server: aai-gateway
Requires Node.js 22.14 or newer.
Use an explicit package selector and binary. @gaia-research/mcp registers two
binaries, so a package name alone cannot select the MCP server:
claude mcp add gaia -- npx --yes --package=@gaia-research/mcp@latest gaia-mcpFor another MCP client, use this command and argument shape:
{
"command": "npx",
"args": ["--yes", "--package=@gaia-research/mcp@latest", "gaia-mcp"]
}The npx-friendly alias is the shortest one-shot command:
npx --yes skill-hell@latest summon "code review" --cardYou can also select the skill-hell binary directly from the rich package:
npx --yes --package=@gaia-research/mcp@latest skill-hell summon "code review" --cardFor a persistent shell installation, install the rich package and use either of its two binaries:
npm install --global @gaia-research/mcp@latest
gaia-mcp
skill-hell summon "code review" --cardskill-hell@latest is an alias package with its own forwarding binary. The
scoped package deliberately exposes both gaia-mcp and skill-hell; always
select the intended binary with --package when running it through npx.
npm ci
npm run build
node dist/bin/gaia-mcp.jsThe server speaks MCP over stdio, so it normally appears idle when run directly.
Connect an MCP client to the absolute path of dist/bin/gaia-mcp.js.
For isolated source testing, override the public projection endpoints with
TREE_URL and TREE_NAMED_URL.
summon materializes the whole skill directory — SKILL.md plus referenced
files, templates, scripts, and fixtures — under an ephemeral session root. It
returns cards, inspect links, timing, cache state, and its ranking disclosure.
skill-hell summon "code review" --count 3
skill-hell sessions
# Re-attach in a new shell/session:
eval "$(skill-hell attach skill-hell-AbCd12)"--count is bounded to 1–5. Named Skills may publish an open trust object;
when there is no comparable signal, the result explicitly reports
relevance-only ordering. This per-invocation ordering is not Hell/Heaven
scoring, routing eligibility, or a content-hash admission policy. Those
features are not shipped.
Warm roots remain under os.tmpdir() for the configured TTL. skill-hell
never writes into the current repository, ~/.claude, or permanent skill
configuration.
npm run check # format check, typecheck, unit/protocol/package tests
npm run test:live # current Gaia public-data contract
npm pack --dry-run # publishable artifact contentsGaia MCP is not:
a second Registry implementation;
a direct Registry or Skill Tree mutation path;
a permanent skill installer or background update daemon;
an implemented Hell/Heaven scoring, routing-eligibility, or content-hash admission system;
the thin Heaven/Summon profile merely because it has a summon tool.
Return an evidence-backed dossier for one generic or Named Skill, including relationships, implementations, trust, sources, and data freshness.
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | Generic id, Named Skill id, or Named catalog reference. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint, openWorldHint, idempotentHint, destructiveHint. Description adds value by specifying the dossier contents (relationships, implementations, trust, sources, data freshness) without contradicting 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?
Single sentence, front-loaded with purpose, no unnecessary words. Every part is informative.
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?
For a tool with no output schema, description adequately lists output contents. Could mention output format or size, but overall complete given low complexity.
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 includes one parameter 'id' with basic description; schema description coverage is 100%. Description does not add further parameter meaning beyond what schema provides.
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 tool returns a dossier for a skill, listing specific contents (relationships, implementations, trust, sources, data freshness). It distinguishes from sibling tools (gaia_search and gaia_status) by focusing on a single skill's detailed inspection.
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?
Description implies usage for getting detailed info on a specific skill but does not explicitly state when to use it over siblings (gaia_search, gaia_status) or provide exclusion criteria.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Find generic and Named Skills in the public Gaia Registry. Returns ranked structured results with trust and source freshness metadata.
| Name | Required | Description | Default |
|---|---|---|---|
| kinds | No | ||
| limit | No | ||
| query | Yes | Task, capability, or skill to find. | |
| tiers | No | ||
| types | No | Skill types to include. Alias of tiers for client compatibility. | |
| minStars | No | ||
| installable | No | Filter by a directly linked, non-blocked SKILL.md source. | |
| contributors | No | ||
| minTrustMagnitude | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true, openWorldHint=true, idempotentHint=true, destructiveHint=false. The description adds behavioral context by noting that results are ranked and include trust/source freshness metadata, which is consistent with annotations. No contradictions.
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?
Two sentences, no unnecessary words. The first sentence immediately states the purpose, the second adds output details. Highly efficient.
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 9 parameters and no output schema, the description provides minimal guidance. It does not explain how to use filters like kinds, tiers, minStars, or what 'trust' and 'source freshness' mean. For a search tool with many options, this is insufficient.
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?
With only 33% schema description coverage (3 of 9 parameters have descriptions), the description does not compensate. It does not explain individual parameters like kinds, tiers, minStars, etc., beyond what the schema already provides. The output mention ('trust and source freshness metadata') does not clarify input parameters.
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 tool finds generic and named skills in the public Gaia Registry, with a specific verb ('Find') and resource. It distinguishes from siblings like gaia_inspect and gaia_status by focusing on search and return of ranked results.
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 use for finding skills but does not explicitly state when to use this tool versus alternatives (e.g., gaia_inspect for details on a specific skill). No when-not-to-use or exclusion criteria are provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Report server version, Registry mode, data-contract compatibility, source freshness, counts, and available tools.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true, openWorldHint=true, idempotentHint=true, and destructiveHint=false. The description adds specific behavioral context (what status information is reported) beyond annotations, without contradiction.
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, well-structured sentence that lists all key aspects of the status report without unnecessary words. It is front-loaded with the verb 'Report' and immediately conveys the tool's function.
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 comprehensive annotations, the description covers the main categories of returned information. It could be slightly more complete by mentioning the output format (e.g., JSON), but overall it is sufficient for the tool's simplicity.
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?
The input schema has no parameters and 100% schema coverage. The description adds full meaning by explaining the tool's output categories, which is essential since the schema provides no guidance.
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 explicitly states the tool reports server version, Registry mode, data-contract compatibility, source freshness, counts, and available tools, clearly defining its purpose as a status check. It distinguishes itself from sibling tools (gaia_inspect, gaia_search) by focusing on a broad overview.
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 usage for obtaining a quick server status, but does not explicitly state when not to use it or point to alternatives. It is clear enough for a simple status tool, but lacks explicit differentiation.
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: gaia_inspect provides detailed information on a specific skill, gaia_search finds skills in the registry, and gaia_status reports server metadata. No overlap or ambiguity.
All tool names follow the consistent pattern 'gaia_<verb>' (gaia_inspect, gaia_search, gaia_status), making it predictable and easy to understand.
With 3 tools, the set is well-scoped for a registry querying server. Each tool serves a necessary function without redundancy.
Covers core query operations (search, inspect, status). Minor gaps like bulk listing or category browsing are absent, but the surface is adequate for typical tasks.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Search & install 6,500+ AI agent skills from skills-hub.ai inside any MCP tool.
AgentPMT is the AI agent marketplace that turns any MCP-compatible AI assistant into an autonomous employee. Connect once and your agents gain access to a growing ecosystem of tools, workflows, and skills spanning communication, data analytics, development, file management, search, and more. AgentPMT dynamically discovers and orchestrates tools from across the MCP ecosystem, so your agents can independently find the right tool for any task without manual configuration.
Governed app access for AI agents: 1,000+ apps & 12,000+ tools via Code Mode MCP.
Governed AI agent skills — one library, distributed to devs and exposed to remote agents over MCP.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/gaia-research/gaia-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
