Brandfetch
by Brandfetch
Pull any company's logos, colours, fonts and brand voice into your AI agent. Brandfetch documents seven read-only tools, OAuth sign-in with a single read scope, a free tier of 100 requests a month, and a fair-use policy you have to read before you ship logos.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Brandfetch via MCP
https://mcp.brandfetch.io/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Brandfetch Tools & Capabilities (7)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- The directory listing is one tool short.
send_feedbackis documented by Brandfetch and absent from Anthropic's 2026-08-16 snapshot. Build from the vendor page, not the listing. - The free tier is 100 requests a month, and MCP calls draw on the same Brand fetch pool as the Brand API, Brand Context API and Transaction API. An exploratory agent session can spend that. Overage billing is not available on the free plan.
- We could not verify tool behaviour ourselves. The endpoint returns 401 to an anonymous handshake, so no
readOnlyHintordestructiveHintannotation was readable. The read-only classification here is Brandfetch's, from its documentation, not ours from the wire. - Cached brand data is licensed for 30 days only, and the licence renews only if you make a new call inside the window. Let it lapse and the terms oblige you to delete what you cached.
- Logo CDN throughput is capped at 1,000 requests per 5 minutes per IP and 2,400 per 5 minutes per customer, with a fair-use ceiling of 500,000 requests a month on the Logo API. Brand API throughput is a sustained 100 requests a second with a 30,000-request rolling 5-minute hard limit.
build_logo_urlsoutput is display-only. Those URLs cannot be downloaded programmatically under the hotlinking policy; onlyget_brand's credentialedsrcURLs are meant for that.- Programmatic scraping of logo images is prohibited and Brandfetch states it may lead to rate limiting or a block.
- You cannot rebuild Brandfetch. Its usage guidelines rule out apps whose only value is re-exposing its logo data, naming an unofficial logo API as a specific example.
- Trademark risk sits with you. Brandfetch licenses the service, not the brands, and disclaims liability for third-party assets.
- No transaction-descriptor redaction. Nothing in the documented surface filters identifiers out of the string before it is sent.
- An unmatched descriptor is final. A
400carryingFailed to enrich transactionwill not change on retry; only a503is worth repeating.
Frequently asked questions
Does the Brandfetch MCP server need a paid plan?
No. Brandfetch's MCP page says a free developer account works, and its Brand API documentation puts the free tier at 100 requests a month. MCP calls draw on the same Brand fetch quota shared by the Brand API, Brand Context API and Transaction API, so a single pool covers all three. Heavy agent use exhausts 100 requests quickly, and overage billing is paid-plan only.
What data does enrich_transaction actually send to Brandfetch?
The raw descriptor text from a bank or card statement, plus a two-letter country code. Brandfetch's Transaction API reference defines the request body as `transactionLabel` and `countryCode`, both required, with `STARBUCKS 1523 OMAHA NE` as its worked example. Descriptor strings can carry store numbers, cities and payment-processor prefixes, so treat this as financial data leaving your environment.
Can I use logos I retrieve through Brandfetch MCP however I want?
No. Brandfetch's terms state third-party brand assets stay the property of their owners, and the licence it grants does not extend to those original logos. Its fair-use guide permits referring, commentary, education and parody, and rules out commercial use on your own products, confusing similarity and trademark damage. Brandfetch also asks no attribution for the Logo API.
Why does get_asset_base64 make my Claude session slow and expensive?
Because it routes the image bytes through the conversation itself. Brandfetch calls the tool a last resort and warns that base64 fallback is slower, can corrupt files and drives up token usage and cost. The documented fix is enabling network egress for code execution and allowlisting `*.brandfetch.io`, so the agent downloads the credentialed `src` URLs from `get_brand` directly.
How long can my agent cache brand data from Brandfetch?
Thirty days from retrieval, and only with a refresh. Brandfetch's terms grant a revocable licence to download, store and cache retrieved Content for a maximum of thirty days, and require a fresh API call at least once every thirty days to renew it. Let the period lapse without that call and the licence expires, obliging you to delete the cached content.
Does the Brandfetch MCP server let an agent change anything?
No write tool touches your account or any brand record. Every documented tool reads brand data, builds a CDN URL, or sends feedback to Brandfetch's own team. The one outbound tool is `send_feedback`, which posts a bug report or feature request. We could not read the server's `readOnlyHint` annotations, because the endpoint returns HTTP 401 to an anonymous handshake.
Can I use Brandfetch MCP without OAuth, in CI or a scripted agent?
Yes. Brandfetch documents an MCP token generated in the Keys and MCP section of its Developer Dashboard, sent as a bearer token in an `Authorization` header. That path exists for clients with no OAuth support and non-interactive environments. Brandfetch tells you to treat the token like a password and keep it out of commits, since it carries the same access as an interactive sign-in.
Sources
- Brandfetch MCP documentation — https://docs.brandfetch.com/mcp/overview (fetched 2026-08-19 as Markdown via the
.mdsuffix) · retrieved 2026-08-19 - Brandfetch Transaction API reference — https://docs.brandfetch.com/reference/transaction-api (fetched 2026-08-19) · retrieved 2026-08-19
- Brandfetch Transaction API overview — https://docs.brandfetch.com/transaction-api/overview (fetched 2026-08-19) · retrieved 2026-08-19
- Brandfetch Brand API overview — https://docs.brandfetch.com/brand-api/overview (fetched 2026-08-19; source of the 100-request free tier and Brand API throughput limits) · retrieved 2026-08-19
- Brandfetch Brand Context API overview — https://docs.brandfetch.com/brand-context-api/overview (fetched 2026-08-19) · retrieved 2026-08-19
- Brandfetch Logo API overview — https://docs.brandfetch.com/logo-api/overview (fetched 2026-08-19; source of the no-attribution statement, hotlinking policy, usage guidelines and CDN rate limits) · retrieved 2026-08-19
- Brandfetch Guide to Logo Fair use — https://docs.brandfetch.com/guides/logo-fair-use (fetched 2026-08-19) · retrieved 2026-08-19
- Brandfetch Terms and Conditions — https://brandfetch.com/terms?mode=dev (fetched 2026-08-19;
docs.brandfetch.com/support/terms-and-conditionsredirects here, and the.mdform is served by Cloudflare as a block page to curl — read with a browser user agent) · retrieved 2026-08-19 - Brandfetch docs machine index — https://docs.brandfetch.com/llms.txt (fetched 2026-08-19) · retrieved 2026-08-19
docs.brandfetch.com/robots.txtdisallows only/cdn-cgi/and/_next/, names a sitemap, and carriesContent-Signal: ai-train=yes, search=yes, ai-input=yes— synthesis is expressly permitted on all three axes (fetched 2026-08-19).developers.brandfetch.comserves a 404 page in place ofrobots.txtandllms.txt, and returns 403 on/docs/mcp; no page on that host was used. · retrieved 2026-08-19- Live RFC 9728 protected-resource descriptor — https://mcp.brandfetch.io/.well-known/oauth-protected-resource (fetched 2026-08-19; the
/mcp-append form returns 200 with only theresourceandresource_metadatavalues differing, and the path-insert form returns 404. Neither body carriesscopes_supported.) · retrieved 2026-08-19 - Live RFC 8414 authorization-server metadata — https://developers.brandfetch.com/.well-known/oauth-authorization-server (fetched 2026-08-19;
scopes_supported: ["read"], PKCES256, dynamic client registration). The MCP host publishes no RFC 8414 document; all three forms return 404. · retrieved 2026-08-19 - Live anonymous probe — 2026-08-19: an anonymous MCP initialize POST to the endpoint returns HTTP 401 with the challenge carried in
x-amzn-remapped-www-authenticate, notwww-authenticate; a bare GET also returns 401 · retrieved 2026-08-19 - Anthropic connector directory — https://claude.ai/directory/76d2c5db-e971-4091-8901-ddb66088956c (snapshot 2026-08-16; six tool names,
prompt_names: [], nopermissionsfield) </content> </invoke> · retrieved 2026-08-16
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Productivity
- Developer
- Brandfetch
- Tools
- 7
- Domain
- mcp.brandfetch.io
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.