Sapiom: The Execution Layer Agents Buy Through
Sapiom announced a $35 million Series A led by Dragonfly on August 5, 2026, bringing total funding to $50 million after a $15 million seed led by Accel in February. Founder and CEO Ilan Zerbib spent five years as a payments director at Shopify, and the thesis reads like it came from that seat: every account, credential, and payment on the internet assumes a human operator, and the new operator is not one.
The company launched as payment rails for agents. It acquired Fewsats in June 2026 to move machine-native payment into the execution layer, and what shipped alongside the Series A is broader than payments: a router, a build environment, and a runtime.
The three products
Router. An OpenAI-compatible endpoint that picks the model for you. Send the same request and Sapiom selects the most efficient allowed model based on task shape, context size, cost, quality, latency, reliability, availability, and company policy, then meters the call. The pitch is spending frontier money only when the task needs frontier reasoning.
agent.studio. A build environment with a coding agent in it, where you move between Canvas, Steps, and Code as the agent takes shape. You describe an outcome and it scaffolds a typed step graph: entry steps, gates, exits. The demo agent collects sources, reads them, writes a cited digest, then pauses at an approval gate before either exit is recorded.
Runtime. Managed execution of those graphs, with explicit retries, signals, schedules, and per-attempt traces. Agents can wait, resume, and stay inspectable, which is the difference between an agent that runs for seconds and one that runs for days.
You can start from a template, bring existing TypeScript, or scaffold from intent. There is a CLI, an MCP server, and a plugin for Claude Code, Cursor, and Codex, so the build loop happens inside whatever coding agent you already use.
What it meters
This is where the product’s shape becomes clear, and it is the question worth asking of any agent platform.
Two things are metered: runs and capability calls.
Runs are the plan. The Developer tier is $0 a month and includes 50 runs a day for accounts created by August 31, dropping to 10 a day after that, with additional runs at $1 each. Startup is $200 a month for 2,000 runs, with additional runs at $0.50. Scale is custom, with committed spend, org-wide policies, audit trail, SSO, SLAs, and SOC 2.
That $1 per run is the number to sit with. A free tier that meters at ten runs a day is a development allowance, not a production one, and an agent firing on real events will pass it in an afternoon.
Capabilities are metered at call time at published unit rates. Sapiom lists AI-powered web search at $0.006 per search, browser extraction at $0.01, image generation at $0.004 per megapixel, audio generation at $0.08, and phone verification at $0.015 per verification. Model access is metered per token. Local runs with stubbed capability calls are free on every tier, so you can validate a graph without paying for what it would have called.
The genuine convenience is one bill. Every capability call is attributed back to its agent and trace, with receipts, and you do not open a vendor account for each service the agent touches. That is the Fewsats acquisition showing up in the product: the agent pays for what it needs mid-run instead of waiting for someone to add a card.
Sapiom says the platform has handled over 270 million transactions and runs more than 100,000 agent runs a day. Those are company figures.
Who it is for
Sapiom is infrastructure for people shipping agentic products. The first-run instructions involve a terminal, the artifacts are typed step graphs, and the customer profile from its seed coverage was vibe-coding platforms whose users should never have to sign up for Twilio and paste an API key.
If you are building a product where agents act on behalf of many end users, the value is real and the alternative is assembling model routing, retries, tracing, and per-vendor billing yourself. If you want an agent for your own work, this is the layer underneath the thing you want, and the run meter will feel strange, because you will be counting runs rather than results.
For that case the shape to look for is an assistant that already watches the things you care about. Carly starts from your inbox and calendar and acts when something arrives, with no graph to author and no runtime to deploy: free Zapier-style workflows, with AI agents from $35/month. Driving it from your own code or provisioning agents per tenant is available too, so if you are building on top rather than using it directly, book a call with the team.
If you are comparing agent infrastructure, Cloudflare’s Agents SDK solves an overlapping problem with a very different cost model, and MCP servers cover the tool-connection layer that both of them lean on.
FAQ
Is Sapiom free?
There is a $0 Developer tier. It includes 50 runs a day for accounts created by August 31, 2026, then 10 runs a day, with extra runs at $1 each. Local runs with stubbed capability calls are always free, and production capability calls are metered at published rates on every tier.
Who founded Sapiom?
Ilan Zerbib, previously a director on payments at Shopify for five years. The company raised a $15 million seed led by Accel in February 2026 and a $35 million Series A led by Dragonfly in August 2026.
What does Sapiom actually charge for?
Runs and capability calls. A run is one execution of an agent graph, priced through the plan with per-run overage. Capabilities like search, browser extraction, image generation, audio, and phone verification are billed per unit when the call executes, and model access is billed per token.
Do I need separate vendor accounts?
No. Sapiom meters each capability call back to its agent and trace and bills it through your Sapiom account, which is the point of the payment layer it acquired with Fewsats.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
