.avif)
Every sales tool on the market now describes itself as AI-powered.
We indexed 1,167 of them and checked something narrower: can an AI agent actually operate this tool through a Model Context Protocol server?
For 224 of them, yes. For the rest, the AI is a feature inside the product rather than something you can point your own agent at.
That gap is the most useful thing in the database, and it is invisible on any vendor's website.
The Sales Tools Database is a free, open directory of 1,167 B2B sales tools, filterable by agentic readiness, MCP server, public API, stack role, pricing model, budget, growth stage, and who the tool is built for. No signup, no gate, updated continuously.
Below: what the numbers say about where the market actually stands, which categories are oversupplied, and how to filter 1,167 tools down to a shortlist.
What you will find in this article
- How many sales tools are genuinely agent-ready, and how many only claim to be
- The 500-tool gap that is about to close
- Which stack role is the most crowded in B2B sales, by a wide margin
- Where the market is thin, and why that matters if you are buying
- The mismatch between tooling and the fastest-growing role in GTM
- How to use the filters to get from 1,167 to five
The agentic gap
Every tool in the database is scored on how ready it is to be operated by an AI agent.
Alongside that, 224 tools ship an MCP server and 722 expose a public API.
Read those together and the picture sharpens. Thirty-eight percent of the B2B sales tool market has no agentic capability at all. Not limited, not partial. None. In a market where nearly every vendor puts AI on the homepage, well over a third of tools cannot be driven programmatically by an agent in any meaningful way.
Meanwhile only 17% are natively agentic, meaning they were built with agent operation as a first-class use case rather than bolted on.
If you are building a stack that an agent will run, that single filter eliminates most of the market before you look at features, pricing, or reviews.
The 500-tool migration wave
Here is the number that predicts the next two years: 722 tools have a public API and only 224 have an MCP server.
That leaves roughly 500 tools that are already programmable but not yet agent-ready. They have the hard part built. The data model is exposed, the endpoints exist, and authentication works. What they lack is the thin protocol layer that lets an agent discover and call those endpoints without a developer wiring each one by hand.
That layer is cheap to add relative to building an API from scratch. Which means most of those 500 will ship MCP servers, and the ones that do not will start losing deals to the ones that do.
Four tools in the database currently list an MCP server as coming soon. Expect that number to look very different by the end of the year.
Two practical consequences for anyone buying now.
A public API is a reasonable proxy for future agent-readiness. If a tool you want has an API but no MCP server yet, it is a far safer bet than a tool with neither.
No API at all is close to disqualifying for anything central to your stack. A tool that cannot be called programmatically in 2026 is a tool you will be migrating off.
Sequencers are the most crowded category in B2B sales
Tools in the database are tagged by the role they play in a stack. Many carry more than one.
273 sequencers. Nearly a quarter of the entire indexed market does essentially the same job: put messages in a cadence and send them.
That number tells you two things. First, sequencing is commoditized, so paying a premium for one is usually a mistake and switching costs should be low. Second, if you are choosing a sequencer, feature comparison will not separate them meaningfully. Deliverability performance, agent-readiness, and integration depth will.
Data sources at 212 and enrichers at 192 make the top three, which together confirm the obvious: most of this market is about finding people and finding out things about them.
Where the market is thin
The bottom of that table is more interesting than the top.
Schedulers: 32 tools. Booking and calendar coordination is a solved-looking problem with surprisingly few dedicated players.
Closers: 48. Almost nothing serves the late stage of the deal compared to the swarm serving the top of the funnel.
Deliverability: 58 tools, against 273 sequencers. There are five times more tools to send email than to make sure it arrives, which mirrors exactly how most teams budget. Our outbound tool stack breakdown makes the same point from the buying side: a third of a working stack should be deliverability, and almost nobody allocates that way.
Verifiers: 77. Also thin relative to the volume of sending the market supports.
If you are building a product rather than buying one, the thin end of this list is where the market is underserved.
The tooling has not caught up to the role
The database tags tools by who they are built for.
SDR and BDR leads at 640 tools. RevOps at 551. Founders at 412. GTM Lead at 332. Account Executives at 280. Marketing at 278. Then GTM Engineer at 239 and Agency at 132.
That GTM Engineer number is low, and it is low for a specific reason. Our GTM Engineering Jobs Report analyzed 739 open postings for a role that barely existed three years ago, with a US median package of $160,000 and only ten junior positions across the entire sample.
The role is growing far faster than the tooling positioned for it. Vendors still segment their marketing around SDRs, because that is who bought software for the last decade. The person actually assembling the stack now is increasingly a GTM engineer, and 239 of 1,167 tools speak to them directly.
For buyers, that means the "best for" filter will under-serve you if you hold that job. Filter on stack role and agentic readiness instead.
The pricing barbell
Two hundred and eighty tools offer a freemium tier, which is the single most common pricing model, ahead of seat-based at 180 and custom at 168. Usage-based sits at 122.
By budget band the distribution is a barbell: 254 tools in the cheapest band, 287 in the second, then a dip to 107 in the third, and back up to 211 in the most expensive.
That dip in the middle is the familiar shape of a market splitting into self-serve tools and enterprise contracts, with relatively little in between. If you are a mid-market buyer, expect to either stitch together several cheap tools or jump to a sales conversation and a custom quote, with fewer clean options at the level in between.
One more detail: usage-based pricing at 122 tools is still a minority model, but it is the one that fits agentic workflows best. An agent running enrichment does not need a seat. Expect that number to rise alongside MCP adoption.
Getting from 1,167 to five
The filters are the product here. A sensible sequence:
Start with stack role, not with category. What job are you actually filling? Sequencer, enricher, verifier, orchestrator. This is the fastest cut and the one most people skip in favor of browsing.
Then filter agentic readiness, and be honest about how you will run this. If a human clicks the buttons, this filter does not matter. If you intend to have agents operate the stack, set it to native or capable and watch the list collapse.
Then API, if agent-readiness alone is too strict. A public API keeps future options open even where MCP is missing today.
Then budget and pricing model. The pricing model matters more than the band. Seat-based pricing on a tool an agent operates makes no sense, and you will feel that within a quarter.
Then company size or growth stage. Mid-market has 901 tools available, scale-up 892, enterprise 451, and solo 303. These are broad filters and work best last.
Five tools you have compared properly beats fifty you have bookmarked.
Five ways people pick tools badly
Choosing the sequencer first. It is the most crowded and most commoditized category in the database. Decide your data and deliverability layers first, then pick something competent to send.
Believing AI claims without checking the protocol layer. Nearly every vendor says AI. Fewer than one in five can actually be driven by your agent. Check for MCP or at minimum an API.
Buying seat-based pricing for automated workflows. If the work is done by software, seats are the wrong unit and you will overpay as you scale.
Ignoring deliverability because there are fewer tools in the category. Thin supply reflects low demand from buyers, not low importance. It is the most under-bought layer in outbound.
Optimizing for features over integration. In an agentic stack, what a tool connects to matters more than what it does alone.
FAQ
Is the database free? Yes. Open directory, no signup, no gate. Updated continuously.
How many tools does it include? 1,167 across 9 categories, 31 subcategories, and 32 features.
How many sales tools have an MCP server? 224, roughly 19% of the indexed market, with a handful more listing one as coming soon.
What does agentic readiness mean? How well an AI agent can operate the tool directly. 204 tools are native, 391 capable, 124 limited, and 448 have no agentic capability.
How many sales tools have a public API? 722, or about 62%. That is the pool most likely to ship MCP servers next.
What is the most crowded category? Sequencers, at 273 tools. Data sources follow at 212 and enrichers at 192.
What is the least served? Schedulers at 32 and closers at 48, with deliverability thin at 58 relative to 273 sequencers.
Can I suggest a tool? Yes, the directory accepts submissions by email.
How is this different from your tool stack PDF? The PDF is prescriptive: three curated stacks at three budgets, telling you what to buy. The database is exploratory: everything that exists, filterable by how you actually work.
What to take away
Two hundred and twenty-four of 1,167 sales tools ship an MCP server. If you are building a stack an agent will operate, that is your real starting universe, not the 1,167.
Thirty-eight percent of the market has no agentic capability at all, despite near-universal AI messaging. Check the protocol layer, not the homepage.
Roughly 500 tools have an API but no MCP server. That gap is the next migration wave, and a public API is a reasonable bet on a vendor's future.
Two hundred and seventy-three sequencers exist. It is the most commoditized layer in outbound, so decide it last and pay the least for it.
Only 239 tools position for GTM engineers, against 640 for SDRs. The tooling market has not caught up with who is actually building these stacks.
We run outbound and CRM automation systems for B2B teams and work with these tools daily. If you want the stack designed and built rather than researched, book a free consultation.
