MCP server for token-efficient documentation search and URL reading across public docs plus Ref-indexed private repositories, PDFs, and documentation sources.
Ref Tools sends search queries and read requests to the Ref API unless a custom `REF_URL` is configured., The server exposes `ref_search_documentation` and `ref_read_url`; OpenAI-style clients can receive compatible `search` and `fetch` tool names., Prompts can steer Ref toward private docs by using the private-source prompt path, so decide which users and clients may access private Ref resources., Protect Ref API keys, avoid placing keys in shareable config screenshots, and revoke keys when a client or workspace is retired., Review fetched URLs before relying on the returned markdown for security-sensitive implementation details.
Privacy notes
Search queries, read URLs, Ref API keys, session ids, client identifiers, result URLs, module ids, and returned documentation snippets may be processed by the Ref service., Private Ref resources can include repository content, PDFs, internal docs, proprietary API references, architecture notes, and vendor documentation., The HTTP transport can receive API keys through headers, query parameters, or encoded config depending on the client path., Redact keys, private URLs, module ids, internal search terms, and returned passages before sharing MCP logs or transcripts.
Author
Ref Tools
Submitted by
oktofeesh1
Claim status
unclaimed
Last verified
2026-06-06
Decision playbook
Review trust signals before you adopt
Signals are present but mixed. Use the checklist below to confirm the source and operational safety for your environment.
Compare context
Selected
0
Current score
63
Baseline
—
Delta
No baseline selected
No major trust-signal divergence detected in the current selection.
Source and provenance checks
Needs review
Confirm ownership and provenance before trusting install instructions.
Source link availableRequired
Open the canonical repository and verify ownership.
Done
Source provenance statusRequired
Marked as source-backed.
Done
Metadata reviewed
No reviewed flag detected in metadata.
Pending
Safety and privacy checks
Complete
Validate risk disclosures before installation or API wiring.
Safety notes presentRequired
Review the listed safety guidance before running commands.
Done
Privacy notes presentRequired
Review data handling notes before connecting accounts or secrets.
Done
Trust level risk gateRequired
Trust level does not block evaluation.
Done
Package and install checks
Needs review
Check package metadata and artifact integrity signals.
Install payload available
Install or copy payload is available for review.
Done
Package verification flag
No package verification flag provided.
Pending
Checksum metadata
No checksum provided for downloaded artifact.
Pending
Compare-driven decision checks
Needs review
Use compare context to validate trade-offs before adoption.
Compare tray has multiple entries
Add at least one more entry to compare trust differences.
5 safety and 4 privacy notes across 6 risk areas. Review closely: credentials & tokens, network access, third-party handling.
6 areas
SafetyNetwork accessRef Tools sends search queries and read requests to the Ref API unless a custom `REF_URL` is configured.
SafetyGeneralThe server exposes `ref_search_documentation` and `ref_read_url`; OpenAI-style clients can receive compatible `search` and `fetch` tool names.
SafetyLocal filesPrompts can steer Ref toward private docs by using the private-source prompt path, so decide which users and clients may access private Ref resources.
SafetyCredentials & tokensProtect Ref API keys, avoid placing keys in shareable config screenshots, and revoke keys when a client or workspace is retired.
SafetyGeneralReview fetched URLs before relying on the returned markdown for security-sensitive implementation details.
PrivacyCredentials & tokensSearch queries, read URLs, Ref API keys, session ids, client identifiers, result URLs, module ids, and returned documentation snippets may be processed by the Ref service.
PrivacyThird-party handlingPrivate Ref resources can include repository content, PDFs, internal docs, proprietary API references, architecture notes, and vendor documentation.
PrivacyCredentials & tokensThe HTTP transport can receive API keys through headers, query parameters, or encoded config depending on the client path.
PrivacyExecution & processesRedact keys, private URLs, module ids, internal search terms, and returned passages before sharing MCP logs or transcripts.
Disclosure: MIT-licensed MCP package for Ref Tools. The server depends on the Ref API and requires an API key, so treat it as a third-party documentation-search service rather than a fully local crawler.
Safety notes
Ref Tools sends search queries and read requests to the Ref API unless a custom `REF_URL` is configured.
The server exposes `ref_search_documentation` and `ref_read_url`; OpenAI-style clients can receive compatible `search` and `fetch` tool names.
Prompts can steer Ref toward private docs by using the private-source prompt path, so decide which users and clients may access private Ref resources.
Protect Ref API keys, avoid placing keys in shareable config screenshots, and revoke keys when a client or workspace is retired.
Review fetched URLs before relying on the returned markdown for security-sensitive implementation details.
Privacy notes
Search queries, read URLs, Ref API keys, session ids, client identifiers, result URLs, module ids, and returned documentation snippets may be processed by the Ref service.
Private Ref resources can include repository content, PDFs, internal docs, proprietary API references, architecture notes, and vendor documentation.
The HTTP transport can receive API keys through headers, query parameters, or encoded config depending on the client path.
Redact keys, private URLs, module ids, internal search terms, and returned passages before sharing MCP logs or transcripts.
Prerequisites
Ref API key for hosted or stdio use.
MCP client with streamable HTTP support, or Node.js and npm for the legacy stdio package.
Agreement on whether the agent may search only public documentation or also Ref-indexed private resources.
Private repositories, PDFs, and docs uploaded to Ref reviewed for sensitivity before enabling private search prompts.
Schema details
Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
10 minutes
Difficulty
beginner
Tool listing metadata
Disclosure
MIT-licensed MCP package for Ref Tools. The server depends on the Ref API and requires an API key, so treat it as a third-party documentation-search service rather than a fully local crawler.
Ref Tools MCP Server gives Claude documentation search and URL-reading tools for
technical work. It can search public documentation and Ref-indexed private
resources, then read selected URLs as markdown so agents can pull focused
context for a specific library, API, service, or framework question.
Use it when a coding agent needs current documentation without stuffing an
entire docs page into context, or when a team has Ref-indexed private
repositories, PDFs, or docs that should be searchable through an MCP client.
These sources were reviewed on 2026-06-06. Prefer the live repository,
README, license, package metadata, MCP registry metadata, manifest, server
implementation, Smithery configuration, and Dockerfile for current setup and
transport details.
Features
Search documentation with ref_search_documentation.
Read a selected URL as markdown with ref_read_url.
Use OpenAI deep-research compatible search and fetch aliases when the
client identifies itself as an OpenAI MCP client.
Search public documentation by default.
Search private Ref resources through the private-docs prompt path.
Run through the hosted streamable HTTP endpoint with an API key.
Run the legacy stdio package with REF_API_KEY.
Support HTTP deployment through the included container and Smithery metadata.
Installation
For the hosted streamable HTTP endpoint, configure the Ref MCP URL with a Ref
API key:
Use the hosted endpoint when the MCP client supports streamable HTTP, and use
stdio when a local MCP package is required.
Use Cases
Ask Claude to search official docs before writing code against a library.
Read only the selected documentation URL that matches a search result.
Search private Ref-indexed repositories, PDFs, or internal docs when the user
is authorized to use them.
Reduce documentation context by searching first and reading the most relevant
page afterward.
Support OpenAI deep-research clients with compatible search and fetch tool
names.
Keep API-reference lookups separate from broad web search.
Safety and Privacy
Ref Tools routes documentation searches and URL reads through the Ref API unless
a deployment overrides REF_URL. Treat search queries, URLs, returned snippets,
session ids, module ids, private-source prompts, and API keys as sensitive. Do
not include private repository names, customer identifiers, unreleased product
details, or proprietary API designs in search prompts unless the Ref workspace is
approved for that data.
The private-docs path can search resources uploaded or indexed in Ref, including
repositories, PDFs, and internal documentation. Review which clients, users, and
workspaces receive the Ref API key, and redact keys, private URLs, internal
search terms, and returned passages from logs or screenshots.
Duplicate Check
No ref-tools/ref-tools-mcp, Ref Tools MCP, Ref MCP, documentation-search MCP,
private-docs MCP, or matching source URL entry was found in content/mcp or
README.md. Existing search and documentation MCP entries do not cover Ref's
API-backed public and private documentation search server.
Show that Ref Tools MCP Server is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/mcp/ref-tools-mcp-server)
How it compares
Ref Tools MCP Server side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
1 trust signal differ across this comparison (Submitter).
MCP server for token-efficient documentation search and URL reading across public docs plus Ref-indexed private repositories, PDFs, and documentation sources.
Hosted streamable-HTTP MCP server registered as ai.com.mcp/skills-search that exposes Agent Skills Search API tools for discovering skills from the skills.sh registry using the open Agent Skills format.
Real-time product search and price comparison across 11M+ products from Singapore, Southeast Asia, and US marketplaces via a remote streamable-HTTP MCP endpoint with API key auth.
Local-first codebase intelligence MCP server that indexes repositories with tree-sitter, stores searchable chunks in DuckDB, and gives Claude semantic search, regex search, daemon status, and deep code research tools.
✓Ref Tools sends search queries and read requests to the Ref API unless a custom `REF_URL` is configured.
The server exposes `ref_search_documentation` and `ref_read_url`; OpenAI-style clients can receive compatible `search` and `fetch` tool names.
Prompts can steer Ref toward private docs by using the private-source prompt path, so decide which users and clients may access private Ref resources.
Protect Ref API keys, avoid placing keys in shareable config screenshots, and revoke keys when a client or workspace is retired.
Review fetched URLs before relying on the returned markdown for security-sensitive implementation details.
✓Search results expose skill metadata and install instructions; installing a skill can give an agent executable scripts, shell commands, and third-party repository access.
Treat discovered skills like unreviewed dependencies—inspect `SKILL.md`, scripts, and publisher sources before `npx skills add` or equivalent installers run.
Self-hosted HAPI deployments inherit API rate limits and outbound network access to `https://skills.sh/api`; scope firewall and logging policies accordingly.
Hosted remotes process search queries on vendor infrastructure; avoid sending confidential project names or customer identifiers in search strings when policy requires on-prem discovery.
✓Read-only product catalog access by default; no write or delete operations exposed.
API key should be stored in the MCP client config or environment, never committed.
✓ChunkHound reads source files, Markdown, text, PDFs, and supported config files under the target directory and stores indexed chunks in a local database.
Realtime indexing and daemon mode can continue watching project files after the initial MCP connection.
Code research and web search tools require embedding, reranking, and LLM configuration and may invoke local CLIs or external model APIs depending on settings.
Exclude generated files, vendored dependencies, secrets, large artifacts, and unrelated repositories before indexing broad workspace roots.
Review MCP client configuration carefully when using an absolute project path in a global Claude Desktop config.
Privacy notes
✓Search queries, read URLs, Ref API keys, session ids, client identifiers, result URLs, module ids, and returned documentation snippets may be processed by the Ref service.
Private Ref resources can include repository content, PDFs, internal docs, proprietary API references, architecture notes, and vendor documentation.
The HTTP transport can receive API keys through headers, query parameters, or encoded config depending on the client path.
Redact keys, private URLs, module ids, internal search terms, and returned passages before sharing MCP logs or transcripts.
✓Search queries, skill names, registry metadata, and tool responses enter MCP client context and may be logged by the hosted remote operator.
skills.sh registry responses can include publisher URLs, install counts, and repository identifiers that reveal technology choices to shared clients.
Installing skills from search results may download repository archives or scripts that process local files; review each skill's privacy posture before use.
Self-hosted HAPI CLI mode keeps search traffic on your infrastructure but still contacts the public skills.sh API unless you mirror the backend.
✓BuyWhere API key is sent as a Bearer token in the Authorization header on every request.
Search queries are sent to BuyWhere's API to retrieve product data; no third-party analytics or telemetry is included in MCP responses.
BuyWhere may log API request metadata (endpoint, timestamp, status) for abuse prevention and billing.
✓Indexed chunks, file paths, symbols, comments, Markdown, PDFs, configuration values, database files, daemon state, and search results can reveal proprietary source code and internal architecture.
Embedding, reranking, LLM, and web search providers may receive code-derived queries or snippets if configured.
Local ChunkHound database files, logs, daemon state, and MCP transcripts may retain code-derived context after the session ends.
Avoid sharing ChunkHound databases, config files with API keys, verbose logs, research outputs, and screenshots from private repositories.
Prerequisites
Ref API key for hosted or stdio use.
MCP client with streamable HTTP support, or Node.js and npm for the legacy stdio package.
Agreement on whether the agent may search only public documentation or also Ref-indexed private resources.
Private repositories, PDFs, and docs uploaded to Ref reviewed for sensitivity before enabling private search prompts.
An MCP client that supports streamable HTTP transport, such as Claude Code, Cursor, Cline, or Windsurf.
Review of which discovered skills the agent may install or execute before enabling autonomous workflows.
Optional Docker or HAPI CLI setup if you need a self-hosted skills-search server instead of the hosted remote.
A BuyWhere API key (free at https://buywhere.ai/api-keys).
An MCP client such as Claude Code, Claude Desktop, Cursor, Windsurf, or any MCP-compatible host.
Network access to https://api.buywhere.ai.
Python 3.10 or newer and the `uv` package manager.
A local repository or workspace you are authorized to index.
ChunkHound JSON config reviewed for database path, excludes, embeddings, and LLM provider settings.
Optional embedding provider credentials for semantic search, or regex-only usage when no embedding key is configured.
Add the streamable HTTP MCP endpoint `https://skills-sh.run.mcp.com.ai/mcp` to your MCP client, then call `searchSkills` to query the skills.sh registry.