AgentDraft vs. Slotflow.
Short answer: Slotflow gets atomicity right and coordination wrong. It books in one API call and returns a 409 on conflict — real credit — but the winner is decided by arrival time alone. First-come-first-served is not coordination: with agents of different importance on one calendar, the wrong agent can win, and there is no priority, no recourse, and no record of who lost or why. AgentDraft adds the arbitration layer — agent priority, holds, a bump window, and an append-only audit trail.
Updated
Slotflow launched in March 2026 as "scheduling infrastructure for AI agents" — one API call to check availability and book. Unlike most of the scheduling APIs agents get pointed at, it actually detects the race: bookings are atomic, and the losing request gets an HTTP 409 instead of a silent double-booking. That puts it closer to AgentDraft's wedge than any other product on our benchmark page's radar, and we'd rather say that plainly than pretend otherwise.
The difference is what happens after the race is detected. Slotflow is strictly first-come-first-served: the request that arrives first wins, full stop. There is no agent identity, no priority, no hold you can take while a human confirms, no bump window, and no audit trail. There is also no email surface — no per-agent mailbox for the inbound message that triggered the meeting in the first place.
Picture two agents on one exec's calendar: the EA agent booking a board-prep session, and a newsletter agent blocking time to draft the weekly send. Under first-come-first-served, if the newsletter agent's request lands 40 ms earlier, it wins. The EA agent gets a bare conflict response, the exec gets the wrong meeting, and nothing anywhere records who lost, to whom, or why. FCFS answers "did two writes collide?" — it never answers "which agent should win?"
AgentDraft answers both, at the storage layer. Every booking carries an agent_priority, and the conflict check is a conditional write — one conditional row per time bucket inside a single DynamoDB transaction — so the check is the write. Holds with a TTL let an agent reserve a slot while a human confirms; a 30-second default bump window lets a higher-priority agent evict a fresh lower-priority commit; the losing agent gets a typed 409 outranked carrying the winner's identity and priority; and every hold, commit, and eviction lands in an append-only audit log. In the open-source collision benchmark, the rank-1 agent won 100.0% of races at p99 112 ms across 500 concurrent attempts, with 0 double-commits.
| Comparison point | AgentDraft | Slotflow |
|---|---|---|
| Core thesis | Coordination layer for N agents on one calendar | "One API call to check availability and book" |
| Atomic booking, 409 on conflict | Yes — storage-level conditional write | Yes — the rare competitor that detects the race |
| Race resolution model | Priority-ranked — the right agent wins, not the fastest | Strictly first-come-first-served |
| Agent identity & priority | First-class — per-agent keys, ranked agent_priority | None — requests are indistinguishable |
| Holds with TTL | Yes — reserve a slot while a human confirms | Not offered |
| Bump window | Yes — 30 s default; fresh commits evictable by higher priority | Not offered |
| What a losing agent learns | Typed 409 outranked with winner's identity + priority | Bare conflict — no winner, no reason |
| Audit trail | Append-only log across holds, commits, evictions, mail, rules | Not offered |
| Agent email / mailbox | Yes — per-agent inbound inboxes | Not part of the product |
| Published collision benchmark | Yes — 100.0% one-winner at p99 112 ms, open source | Not published at time of review |
| Pricing entry point | Free Developer tier (no card — 1 agent, 1 mailbox, 50 bookings/mo); $10/mo Individual; $25/mo Team | Free for 100 bookings; then $79/mo, up to $299/mo |
Slotflow rows reflect the public product surface as verified on 2026-07-01. If we've mis-stated a feature, tell us and we'll correct it. We have not run an equivalent benchmark against Slotflow; if their team would like to collaborate on a head-to-head methodology, we'd welcome it.
If exactly one agent writes to the calendar, that agent never needs an email inbox, and "check availability, book the slot" is the whole workflow — Slotflow covers it. With a single writer, first-come-first-served has no one to be unfair to, and the atomicity Slotflow does provide is genuinely the hard part of that narrow problem. The main caveat is price: after the free 100 bookings, Slotflow starts at $79/mo, where AgentDraft's paid plans start at $10/mo and the Developer tier stays free.
- The second agent arrives. The moment two agents share a calendar, "first to arrive" starts deciding questions it has no business deciding. An exec's EA agent and a newsletter-scheduling agent are not peers; FCFS treats them as identical.
- The wrong agent wins and there's no recourse. Slotflow has no bump window and no eviction, so a low-stakes booking that lands first simply keeps the slot. On AgentDraft, a higher-priority agent can reclaim it within the 30-second bump window, and the eviction is recorded.
- Someone asks "why is this meeting here?" With no audit trail and no agent identity, there is no answer. AgentDraft's append-only log names the agent, the priority, the losers, and the timestamps for every slot.
- The workflow touches email. Bookings are usually triggered by mail and confirmed by mail. Slotflow has no mailbox surface; every AgentDraft agent gets its own addressable inbox.
Frequently asked
Doesn't Slotflow already prevent double-bookings?
Yes — and we say so plainly: Slotflow books atomically and returns a 409 on conflict, which is more than most scheduling APIs give an agent. But preventing a double-booking is not the same as resolving a conflict well. Slotflow's winner is whichever request arrives first, regardless of how important the agent behind it is. AgentDraft ranks agents, lets a higher-priority booking bump a fresh lower-priority one inside a 30-second window, and records every outcome in an append-only audit log.
When is Slotflow enough on its own?
When exactly one agent writes to the calendar, the agent never needs an email inbox, and check-and-book is the whole workflow. With a single writer, first-come-first-served has no one to be unfair to. Note the price, though: Slotflow is free for 100 bookings and then starts at $79/mo, while AgentDraft has a free Developer tier and paid plans start at $10/mo.
What does a losing agent get from AgentDraft that it doesn't get from Slotflow?
A typed 409 that says why it lost — outranked — and carries the winning agent's identity and priority, plus an audit reference. On Slotflow, a losing agent gets a bare conflict response: no winner, no reason, no recourse. Retry logic can only do the right thing when it knows whether it lost to a higher-priority agent or to a plain race.
How do the free tiers compare?
Slotflow's free tier covers 100 bookings; after that it's $79/mo, rising to $299/mo. AgentDraft's Developer tier is free with no card required — 1 agent, 1 mailbox, and 50 bookings per month — and paid plans start at $10/mo for 3 agents, or $25/mo for a team with shared priority resolution.
- The AgentDraft collision benchmark — the 500-attempt receipts behind the comparison.
- Slotflow alternatives — the wider field, if you're comparing more than these two.
- The agent-scheduling glossary — bump window, hold TTL, outranked 409, and the rest of the vocabulary.
- AgentDraft vs. Cronofy — the related comparison if you're also evaluating provider-layer APIs.