Nemo Retriever MCP

From the author

Use when a task needs to search or add documents through NeMo Retriever MCP.

— the description by NVIDIA

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The NeMo Retriever MCP skill provides a structured interface for interacting with the NeMo Retriever service via the Model Context Protocol (MCP). It enables agents to perform semantic searches across indexed documents and ingest new content when necessary. The skill emphasizes using specific tool calls for querying and ingestion while ensuring that agents properly ground their responses in the retrieved evidence, including maintaining document identifiers for accurate citations.

What it does

Provides access to NeMo Retriever MCP tools for document retrieval and ingestion.
Supports querying existing document indices to find relevant information.
Facilitates the ingestion of new documents into the retriever system.
Guides the agent to ground answers using retrieved evidence hits.
Ensures document metadata and identifiers are preserved for citation purposes.

When to use it

When an agent needs to search through a collection of documents to answer a user query.
When new documents must be added to the retriever index for future searchability.
When the task requires evidence-based responses with specific source citations.
When working within an environment that supports NeMo Retriever MCP server tools.

Advantages

Standardizes the interaction pattern for document retrieval tasks.
Promotes accurate grounding by requiring evidence hits to be read and cited.
Prevents redundant ingestion of already searchable documents.
Provides clear error handling and retry guidance for query operations.

Usage examples

“Search for information regarding the company's Q3 policy in the retriever.”: The agent uses the query tool with a search string, retrieves the top 5 hits, and synthesizes an answer citing the specific document IDs and page numbers provided.
“Ingest the document at /shared/data/policy.pdf and then find the section on remote work.”: The agent calls ingest_documents for the specified path, confirms success, and then performs a query to extract the requested information about remote work.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/nemo-retriever-mcp/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip nemo-retriever-mcp.zip -d .agents/skills

Documentation by NVIDIA

Use the NeMo Retriever MCP tools exposed under retriever instead of the retriever CLI.

The server exposes MCP tools, not MCP resources. Do not call list_mcp_resources to discover documents or decide whether Retriever is available; an empty MCP resource list is expected.

Workflow

Use query to search documents already available to Retriever.
Never ingest documents that are already searchable, including documents in a prebuilt index.
Use ingest_documents only when required documents are not yet searchable and server-readable paths are explicitly available. Do not assume that paths visible to the agent are also readable by the Retriever server.
After ingestion completes, use query to retrieve evidence from the added documents.
Start with top_k=5 and format="hits". Request more only when the first successful response does not contain enough evidence to answer the task.
Read evidence hits from the tool response and ground your answer in them.
Preserve available document identifiers such as document_id, source_id, or pdf_basename, along with page numbers, when the task requests citations or structured output.
If query fails, inspect the error and retry with corrected parameters or a shorter equivalent query. If it still fails, report the error before using another retrieval method.

Files

ATTRIBUTION.mdAdded by AutoGineer835 B
BENCHMARK.md7.5 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md1.5 KiB
evals/evals.json4.4 KiB
skill-card.md3.9 KiB
skill.oms.sig4.5 KiB

skill-card.md

Description:

Use when a task needs to search or add documents through NeMo Retriever MCP.

This skill is ready for commercial/non-commercial use.

Owner

NVIDIA

License/Terms of Use:

Apache 2.0

Use Case:

Developers and engineers use this skill to search and ingest documents through NeMo Retriever MCP tools for retrieval-augmented generation workflows.

Deployment Geography for Use:

Global

Requirements / Dependencies:

Requires API Key or External Credential: [Not Specified] Credential Type(s): [None identified]

Do not include secrets in prompts/logs/output; use least-privilege credentials; rotate keys as appropriate.

Known Risks and Mitigations:

Risk: Review before execution as proposals could introduce incorrect or misleading guidance into skills. Mitigation: Review and scan skill before deployment.

Reference(s):

NeMo Retriever Library Documentation
Agentic Retrieval (Concept)
Workflow: Agentic Retrieval

Skill Output:

Output Type(s): [Analysis, API Calls] Output Format: [Markdown with citations] Output Parameters: [1D] Other Properties Related to Output: [None]

Evaluation Agents Used:

Claude Code (aws/anthropic/bedrock-claude-opus-4-8)
Codex (openai/openai/gpt-5.5)

Evaluation Tasks:

Evaluated against 4 tasks (3 positive, 1 negative) with 3 attempts per task in isolated k8s-sandbox pods.

Evaluation Metrics Used:

Reported benchmark dimensions:

Security: Checks for unsafe operations, secret leakage, and unauthorized access.
Correctness: Checks final-answer correctness against the reference answer.
Discoverability: Checks whether the expected skill was selected and the workflow executed.
Effectiveness: Checks whether the user's goal was achieved and expected workflow behavior was followed.
Efficiency: Checks tool-call productivity and token efficiency.

Underlying evaluation signals used in this run:

security: Unsafe operations, secret leakage, and unauthorized access.
skill_execution: Whether the expected skill was selected, decoys were avoided, and the workflow executed.
accuracy: Final-answer correctness against the reference answer.
goal_accuracy: Whether the user's goal was achieved.
behavior_check: Whether the expected workflow behavior was followed.
skill_efficiency: Tool-call productivity.
token_efficiency: Actual uncached prompt plus completion token usage.

Evaluation Results:

Measure	Claude Code (Baseline → Skill Uplift)	Codex (Baseline → Skill Uplift)
Overall	77.1%	69.2%
Security	93.8% → 100.0% (+6.2 points)	30.0% → 90.0% (+60.0 points)
Correctness	17.5% → 60.0% (+42.5 points)	16.0% → 44.0% (+28.0 points)
Discoverability	88.3%	75.0%
Effectiveness	39.3% → 50.6% (+11.3 points)	28.5% → 44.3% (+15.8 points)
Efficiency	86.4%	92.9%

Skill Version(s):

e971432a (source: git SHA, committed 2026-09-11)

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal team to ensure this skill meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

(For Release on NVIDIA Platforms Only) Please report quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Security

What our scanner noticed· Medium risk

Large base64-encoded blobs

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Build MCP App
AnthropicOfficial
Build interactive UI widgets for MCP servers to render forms, pickers, and dashboards directly within the chat interface.
Automation and integrationsClaude CodeClaude appsOpenAI Codex+14 more
Apache-2.0GitHub stars of the source repository46.3k
MCP Integration
AnthropicOfficial
Comprehensive guide for integrating Model Context Protocol (MCP) servers into Claude Code plugins.
Automation and integrationsClaude CodeClaude appsOpenAI Codex+14 more
Apache-2.0GitHub stars of the source repository46.3k
Social Publishing
Seth Hobson· Community
Programmatically schedule and publish content across 13 social media platforms using the SocialClaw API.
Automation and integrationsClaude CodeClaude appsOpenAI Codex+14 more
MITGitHub stars of the source repository42.3k
Hermes Tweet
Seth Hobson· Community
Integrates X/Twitter research and action capabilities into the Hermes Agent framework with strict approval-gated security.
Automation and integrationsClaude CodeClaude appsOpenAI Codex+14 more
MITGitHub stars of the source repository42.3k
Paypal Integration
Seth Hobson· Community
A guide to integrating PayPal payment processing, including checkout flows, subscriptions, and IPN webhook handling.
Automation and integrationsClaude CodeClaude appsOpenAI Codex+14 more
MITGitHub stars of the source repository42.3k
Social Publishing Publisher
Seth Hobson· Community
An agent-driven social media publishing specialist for multi-platform campaign management via the SocialClaw API.
Automation and integrationsClaude Code
MITGitHub stars of the source repository42.3k

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download nemo-retriever-mcp.zip

Nemo Retriever MCP

From the author