Head to head · Music search · October 2026 research run

Free Music Library MCP vs Lots of Sounds

Lots of Sounds and Free Music Library MCP score within a point of each other on agent readiness, 52.9 (D) and 52.4 (D). Free Music Library MCP leads on payments & pricing. Both do music search.

Best music and sound libraries for AI agents · All 21 music libraries comparisons

Which one, for what

Free Music Library MCP D

Best for An agent finding free background music, loops or sound effects for a video, podcast or game, with attribution it can hand to the user.

Ahead on

  • Payments & pricing, 90 against 25

Also in its favour

  • No key needed to call it
  • Runs on your own machine
  • Open source

Watch for

Telemetry is on by default and sends the intent argument verbatim and parent process names, while the README says no user data is collected

Lots of Sounds D

Best for An agent that needs short CC0 sound effects, ambience or loops with trim points and loudness for video, apps or games.

Ahead on

  • Schema & documentation, 79 against 65
  • Agent ergonomics, 88 against 82
  • Maintenance & community, 66 against 31
  • Transparency & trust, 43 against 35

Also in its favour

  • A hosted endpoint, with nothing to install

Watch for

No page says where the CC0 sounds come from or who recorded them, and the licence page leaves checking suitability to the user

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Payments & pricing Free Music Library MCP +65

Free Music Library MCP 90

Free, self-hosted software under MIT with nothing to buy, read with the self-hosted rule. No payment protocol (0), nothing to price (30), free (30), and three sources work with no signup or key (30). Jamendo and Freesound need the user's own free keys. 0 + 30 + 30 + 30 = 90.

Lots of Sounds 25

No x402, MPP or L402 (0). Plan prices are public without login, Pro $15 and Enterprise $49 a month, with no per-call price (15). The free plan needs no card but has no API access. Twelve curated samples answer with no key through browse_samples and /api/v1/sounds/sample, which shows the response shape and nothing more (5 of 30). Full access needs a person to sign up and subscribe in a browser, and the CLI's device-code login also needs a person to approve. The keyless sample tool earns a little (5 of 30). 0 + 15 + 5 + 5 = 25.

Maintenance & community Lots of Sounds +35

Free Music Library MCP 31

Version 0.4.3 published to PyPI on 3 October 2026, 50 days before this check, per the PyPI release feed and the v0.4.3 tag (20). Two releases in the last 90 days, 0.4.2 (20 September) and 0.4.3 (3 October) (0). One maintainer, 43 commits from 3 August to 3 October 2026, and no commits since. Issues and pull requests on GitHub couldn't be read on 22 November (3 of 25, unchecked). server.json names io.github.surendranb/music-mcp, but the official registry lists six other io.github.surendranb servers and not this one, and the PyPI package it names (music-mcp) doesn't exist (0). CI, CodeQL, monthly Dependabot, a uv.lock and a current MCP SDK range (mcp>=2.0.0,<3). Run results unread (8 of 10). 20 + 0 + 3 + 0 + 8 = 31.

Lots of Sounds 66

Closed service. The npm CLI 0.1.0 was published on 29 October 2026, 24 days ago (30). In the last 90 days the examples repository changed on 26 August and 21 September, the registry entry and TypeScript client appeared on 21 September, the privacy policy took effect on 19 October and the CLI shipped on 29 October, but none of these is a dated changelog entry (10 of 20). The blog's only changelog is March 2026, the roadmap lists a shipped MCP server as in progress, and support is team@lotsofsounds.com with a reply within one business day stated on the pricing page and a UserJot feedback board (6 of 15, closed-service reading). Listed in the official MCP registry as com.lotsofsounds/mcp, a reverse-DNS namespace on its own domain (15). The CLI has no install scripts and the TypeScript client has CI, but the CLI's source repository named on npm isn't public (5 of 10). 30 + 10 + 6 + 15 + 5 = 66.

Schema & documentation Lots of Sounds +14

Free Music Library MCP 65

Read from music_mcp/server.py in a clone of the repository. Every tool's inputs come from typed Python signatures, so tools/list carries a JSON Schema for each (25). The repository has llms.txt, AGENTS.md and a Markdown error skill, and the docs pages on builditwithai.xyz describe each tool. No llms.txt was found served for the site (6 of 10). The description passed to search_music is one line, but its docstring and the server instructions say when to read the error skill, what sources and intent take and that attribution must be shown verbatim (11 of 20). limit is clamped to 1 to 50 in code but not in the schema, and sources is a free list of strings with no enum (7 of 15). The interpreting-errors skill documents each error shape and its recovery, and three prompts show worked queries. The README's tool table names five tools the server doesn't have (11 of 15). The server reports its version, and releases are tagged, but there is no changelog (5 of 15). 25 + 6 + 11 + 7 + 11 + 5 = 65.

Lots of Sounds 79

Public OpenAPI 3.1 spec with 7 operations, and typed JSON Schema inputs on all 5 MCP tools, read with one tools/list (25). llms.txt and llms-full.txt, plus a Copy Markdown option on each docs page (10). Tool descriptions say how search matches, when to call download_sound and that get_profile is only for identifying the account, with fewer when-not-to-use lines in the spec (14 of 20). Enums for category, sort and order, minimum and maximum on limit and page, required ids, and additionalProperties: false on every tool. Duration filters have no bounds (13 of 15). Examples in curl, JavaScript and Python and an error table with actions, but the spec's error schema is a single error string and the coded errors (plan_required, rate_limited) are documented for the CLI only (12 of 15). Paths are versioned /api/v1 and the spec is 1.0.0, but the one dated changelog post is from March 2026 (5 of 15). The spec says 12,000+ sounds where llms.txt says 16,874, and tags use OR on the sample endpoint and AND on search. 25 + 10 + 14 + 13 + 12 + 5 = 79.

Transparency & trust Lots of Sounds +8

Free Music Library MCP 35

MIT in LICENSE and pyproject.toml, copyright Surendran B, but every Python source file carries an SPDX-License-Identifier: Apache-2.0 header (25 of 30). No privacy policy. The README says no search queries or user data are collected and SECURITY.md that telemetry never logs search terms or queries. The code sends the model-written intent argument verbatim after scrubbing URLs, paths and emails, plus parent process names from ps. A design note in the repository names PostHog as the destination and says the gateway adds country and city. The statements and the code disagree (8 of 30). No deprecation policy or dated notices (0). Telemetry is on by default and disclosed in the README with working opt-outs (DO_NOT_TRACK, MUSIC_MCP_TELEMETRY=false), but one documented opt-out does nothing and the disclosure understates what is sent (8 of 20). 25 + 8 + 0 + 8 = 41.

Lots of Sounds 43

Closed service with terms of 1 April 2026 and a licence page that put every sound under CC0 1.0. No page read says where the sounds come from, who recorded them or how their CC0 status was checked. Sound ids carry an fs- prefix, sound pages show added dates from 2017, and the roadmap plans generated sounds, so the licence's basis can't be traced (10 of 30). The privacy policy of 19 October 2026 lists what is collected, including search terms, and says data is kept 'as long as reasonably necessary', with no periods, DPA or deletion route beyond an email (10 of 30). No deprecation policy or dated notices (0 of 20). Processors are named by type only (hosting, authentication, payments), with no list or locations (3 of 20). 10 + 10 + 0 + 3 = 23.

Agent ergonomics Lots of Sounds +6

Free Music Library MCP 82

Five compact tools (search_music, list_sources, skills_list, skill_read, check_for_updates) plus three prompts and skill resources (25). limit up to 50 and a sources filter, with no paging or offset (12). A missing key comes back as a skipped entry with key_required and the variable to set, an upstream failure as reason: error, and an unknown source as a named error, all described in the bundled skill. Clients that support elicitation are asked for a missing key mid-call (16). Every tool carries read_only_hint and idempotent_hint, and a search can be repeated safely (20). Only query is required. One language, Python, with the npm package only a launcher. Tool results can carry an appended upgrade notice once a week when PyPI has a newer version (9 of 15). 25 + 12 + 16 + 20 + 9 = 82.

Lots of Sounds 88

Five compact tools, about 5,000 characters of definitions (25). Search takes page and limit (up to 100), category, tags, duration bounds and sort, and returns total and totalPages (20). Errors are documented with a status, a meaning and an action, and 429 carries a reset time, though MCP error shapes were not seen because no tool was called (15 of 20). Every tool carries readOnlyHint, destructiveHint and idempotentHint, the reads are idempotent, and download_sound is marked non-destructive (20). Defaults on every optional parameter and one required field at most. The npm CLI and examples in JavaScript and Python exist, but the TypeScript client was not found on npm, so there is one published package (8 of 15). 25 + 20 + 15 + 20 + 8 = 88.

Score by category

CategoryWeightFree Music Library MCPLots of SoundsEdge
Reliability16%4340Free Music Library MCP +3
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%6579Lots of Sounds +14
Agent ergonomics13%8288Lots of Sounds +6
Security & auth14%4546Lots of Sounds +1
Payments & pricing10%9025Free Music Library MCP +65
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%3166Lots of Sounds +35
Transparency & trust7%3543Lots of Sounds +8
Negative events≤15-5-3
TotalD 52.4/100D 52.9/100

Facts side by side

FactFree Music Library MCPLots of Sounds
KindMCP serverMCP server
VendorBuildItWithAI Studio (Surendran B.)Lots of Sounds (Elliott Labs)
Hosted endpointno (local only)https://api.lotsofsounds.com/mcp
TransportsstdioStreamable HTTP
AuthNoneOAuth or key
PricingFreeFreemium
x402nono
LicenceMITnone
Tools exposed55
Read-only variant documentednoyes
llms.txtnoyes
MCP registrynot listedcom.lotsofsounds/mcp
Last release2026-10-032026-10-29
Terms last updatedno document linked2026-04-01
Privacy policy last updatedno document linked2026-10-19
Customer content may train modelsnot read yetnot found in the text
Terms restrict automated accessnot read yetnot found in the text
Terms restrict benchmarkingnot read yetnot found in the text
Terms or service can change without noticenot read yetnot found in the text
Arbitration or class-action waivernot read yetnot found in the text

Verdicts

Free Music Library MCP

Five read-only tools search five free music sources and return each track's licence and attribution line, three of them with no key. Telemetry is on by default, sends the model's stated intent verbatim, and the documented MCP_TELEMETRY_OPT_OUT does nothing. The installer and the update notice point to package names nobody has published.

Lots of Sounds

A small, read-only MCP server and REST API with typed inputs, tool annotations, numbered daily limits and timing and loudness data on each sound. API access needs a person to sign up and subscribe, no page says where the CC0 sounds come from, and the home page's brand logos and uptime SLA have nothing behind them.

Before you call either

Free Music Library MCP

  1. Set DO_NOT_TRACK=1 or MUSIC_MCP_TELEMETRY=false in the server's env to stop telemetry. MCP_TELEMETRY_OPT_OUT=1 has no effect
  2. Install with uvx free-music-library-mcp rather than the curl installer, which edits every MCP client config it finds
  3. If a result says to run uvx --refresh music-mcp, use uvx --refresh free-music-library-mcp instead, since music-mcp is not this package on PyPI
  4. Pass the attribution string to the user verbatim, and drop CC BY-NC tracks when the use is commercial
  5. Treat a skipped entry with key_required as final for that source and search the keyless sources instead of retrying

Lots of Sounds

  1. Send short keywords in q, because every word must match a name, description or tag and search isn't semantic
  2. Call download_sound only when the user is ready, because the signed URL expires after one hour
  3. If the server returns 429, wait for X-RateLimit-Reset, which falls at midnight UTC for the daily limit
  4. Pass one or two tags on search_sounds, since tags match with AND logic there and with OR logic on the sample endpoint
  5. Tell the user the source of each sound isn't stated before they use it in a product where provenance matters

Questions

Which is better for AI agents, Free Music Library MCP or Lots of Sounds?

Lots of Sounds and Free Music Library MCP score within a point of each other on agent readiness, 52.9 (D) and 52.4 (D). Free Music Library MCP leads on payments & pricing.

Do Free Music Library MCP and Lots of Sounds need an API key?

Free Music Library MCP needs no key. Lots of Sounds takes an API key or an OAuth sign-in.

Can an agent call Free Music Library MCP and Lots of Sounds without installing anything?

Free Music Library MCP runs on your own machine, with no hosted endpoint listed. Lots of Sounds has a hosted endpoint at https://api.lotsofsounds.com/mcp.

Are Free Music Library MCP and Lots of Sounds open source?

Free Music Library MCP is open source (MIT). No open-source release is listed for Lots of Sounds.

Other comparisons with Free Music Library MCP or Lots of Sounds

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.