Brandfetch MCP Server: Real Brand Logos and Colors for AI Agents

An agent asked to put a company's logo on a slide will draw one from memory, and memory is not a design asset. It gets the shape approximately right, the colors close but not the hex values, and the result reads as fake because it is fake. Brandfetch MCP is a small, specific fix for that problem: an MCP server that gives any agent a direct lookup into Brandfetch's brand-data platform, covering more than 50 million companies, so it can pull a real logo file, a real color palette, and a real font stack instead of inventing one.

What The MCP Server Actually Exposes

The server, documented at docs.brandfetch.com/mcp/overview, ships five tools. brand_search resolves a company from a loose name or partial match. get_brand is the core lookup, returning logos, colors, fonts, and company details for a given domain, stock ticker, ISIN, or crypto symbol. get_brand_context is aimed specifically at content generation: it returns brand voice, positioning, audience, and stylistic notes formatted for an LLM to work from, not just visual assets. enrich_transaction resolves a merchant from a raw payment descriptor string, useful for finance and expense-tracking agents. build_logo_urls generates ready CDN logo URLs without spending an API call.

That combination matters more than it first looks. Most "logo API" tools stop at the image. Brand context and transaction enrichment push this toward general brand-lookup infrastructure for agents doing research, content drafting, or financial categorization, not just designers filling a template.

Connection is via OAuth: an MCP client opens a browser window, you sign into a Brandfetch account, and the server runs at mcp.brandfetch.io/mcp. For non-interactive setups, scripted agents or CI, you generate a bearer token in the developer dashboard instead. It plugs into Claude, Cursor, VS Code, and other MCP-compatible clients the same way any other server does.

Traction, Verified

Brandfetch's own product page on Product Hunt carries multiple launches stacked under one listing, going back to 2019, which is exactly the distortion that made the scan's first pass unreliable. Checking the actual launches list resolves it cleanly this time: "Brandfetch MCP" is its own distinct sub-launch, dated August 6, 2026, sitting at 136 upvotes under the tagline "Stop your AI from guessing brand logos." That is a real, current number attributable to the MCP launch itself, not inherited from the parent product or an older release. For context, Brandfetch's original Brand API launch in 2022 sits at 252 upvotes and its Personalization API from 2019 sits at 368, so 136 is a solid but unspectacular debut relative to the company's own history, not a runaway hit.

Free Tier and Pricing, Precisely

This is the part worth reading before wiring an agent to it. MCP requests draw from the same quota as the Brand API, and the free tier is 100 requests per month, tracked in the developer dashboard. That is workable for testing or light personal use, thin for anything running unattended or at scale. Paid usage is metered rather than seat-based: overage runs about $0.10 per additional brand call, and a $99-per-month tier raises the cap to roughly 2,500 calls a month. Separately, Brandfetch's Logo API and Brand Search API have a far more generous allowance, up to 500,000 requests a month without attribution, but that generosity applies to raw logo image delivery and brand search, not to the richer get_brand and get_brand_context lookups the MCP server leans on. Budget for the Brand API's tighter free tier, not the Logo API's looser one, when estimating what an agent workflow will actually cost past the first month.

The Part That Actually Matters: Logo Usage Rights

Fetching a real logo file is not the same as having permission to publish it, and this is where a lot of teams get it wrong by assuming an API response equals a license. Brandfetch's own fair-use guidance, published at docs.brandfetch.com/guides/logo-fair-use, is direct about the boundary: logos can be used for reference, identification, commentary, or educational purposes, the kind of thing a review site or a comparison table does when naming which tools it covers. What is not covered is commercial exploitation, meaning putting another company's logo on your own product, merchandise, or marketing material in a way that could be read as endorsement or affiliation. Brandfetch's own test for whether a use crosses that line is worth repeating directly: does the use suggest the product originates from the logo owner, and would it diminish the owner's market position. If either answer is yes, it is very likely not fair use, API access notwithstanding.

Practically, that means a fetched Brandfetch logo is fine for things like an internal dashboard naming vendors, a "works with" style integration list, or an AI-generated blog graphic that is clearly editorial and not implying partnership. It is not a green light for slapping a competitor's mark on your own landing page, an ad, or a product screenshot designed to imply affiliation. The MCP server solves the hand-drawn-approximation problem cleanly. It does not solve the licensing question, and that question does not go away just because the fetch was automated.

Where It Fits

Brandfetch MCP is a narrow, well-scoped tool rather than a platform play, and that is the right shape for what it does. It is not a competitor to logo generation tools like the AI logo generators covered elsewhere on this site, which create new marks for brands that do not have one yet. This does the opposite job: retrieving assets that already exist and are already correct, for brands that already have them. Any agent workflow that touches real companies by name, brand comparison content, vendor directories, competitive research briefs, financial transaction categorization, has a genuine use for a tool that stops it from guessing. The free tier is enough to confirm that fit before deciding whether the metered pricing is worth it for a production workflow, and the fair-use boundary is narrow enough that it is worth reading before the first real asset ships anywhere public.