Plain

What it is

Plain is a B2B support platform you assemble rather than configure. Threads, customers, labels, SLAs and the AI queue are all reachable through a public GraphQL API at core-api.uk.plain.com/graphql/v1, with a typed SDK and a documented MCP server so an agent can work the queue directly. The support UI Plain hands you is one client of that API; your own product can be another.

That inversion is the product. Pylon and Zendesk sell a support application and expose an API around its edges. Plain sells the engine and treats the application as replaceable — the vendor’s own term for this is support infrastructure.

Channels: Slack (including Enterprise Grid and community servers), Microsoft Teams, Discord, email, live chat, a customer-facing support portal, and structured in-app contact forms. Issue trackers: Linear, Jira, GitHub Issues, Shortcut. CRM sync: HubSpot, Salesforce, Attio. Incidents: incident.io and Rootly.

The company raised a $15M Series A led by Battery Ventures in February 2025, with Index Ventures and Connect Ventures returning from a $6M seed. Plain’s published customer list is almost entirely developer-tools companies: Vercel, Cursor, n8n, Fly.io, Clerk, Raycast, Granola, Sanity, Stytch, Prisma and Ashby.

Why it shows up in Customer Success stacks

• Bring your own agent is the feature nobody else in the category sells. Plain lets you attach Claude, GPT, Gemini, or a model you fine-tuned yourself as a first-class participant in the queue — it picks up threads and your humans see its work in the same timeline. Every other vendor here sells you their model and prices the deflection it produces. Plain sells the seat and lets you own the inference bill and the prompt. If your team already runs an internal model with retrieval over your own docs, this is the only support platform in the catalog that will let it answer a customer.
• Two AI products, deliberately split. Ari is the customer-facing agent that deflects. Sidekick is the internal copilot that reads a thread plus your connected Linear, GitHub and HubSpot context and answers the messier question for the human handling it. They bill on different meters, which is the single most important thing to understand before signing.
• Third-party agents are integrations, not competitors. Decimal, Duckie AI, Inkeep and Parahelp all plug in as agents. A platform that publishes its rivals as first-class integrations is telling you where it thinks its moat sits, and it is not the model.
• The custom support surface is the scoped use case it wins outright. Provisioning a dedicated Slack channel per enterprise trial, opening a thread automatically from a product error spike, rendering live ticket status inside your own dashboard, composing per-segment SLAs that match contract value instead of a vendor’s pricing tiers — with Plain these are things you build in an afternoon. With a packaged helpdesk they are feature requests.

Pricing reality

Published, self-serve, and unusually legible for this category — which itself is a buying signal in a segment where Pylon has pulled its rate card behind a demo.

• Foundation — $35/month. One seat included, $35 per additional seat, capped at 5 seats. 2,000 Sidekick credits per month. 7-day free trial, no card.
• Horizon — $299/month. Three seats included, $99 per additional seat, capped at 10 seats. 15,000 Sidekick credits per month.
• Frontier — custom. Custom credit allowance, custom AI and bring-your-own-agent, dedicated CSM and onboarding.

Annual billing saves 10%. Eligible early-stage startups get up to 50% off.

The real-world band: a 4-person support team on Foundation pays $140/month, about $1,512/year billed annually. The same team grown to 8 is forced onto Horizon at $794/month — $299 plus five seats at $99 — or roughly $8,575/year. Plan against the second number, because the step between them is sharp: moving from 5 seats to 6 takes you from $175/month to $596/month, a 3.4x jump for one hire.

Effective seat cost on Horizon settles near $99 per seat per month at any headcount from 4 to 10. That sits above Pylon’s last-published $89 Professional seat and below its $139 Enterprise seat — before either vendor’s AI meter is counted.

The AI meter is where Plain is genuinely cheap. Ari is currently free on every workspace, so customer-facing deflection carries no per-resolution charge at all. Zendesk bills $1.50 per automated resolution on a commitment or $2.00 pay-as-you-go. Pylon’s AI Agents start at $100/month and scale with issue volume. At 2,000 deflected conversations a month, that is roughly $3,000/month of line item Plain does not currently charge for.

Credits are the unpredictable half. They meter Sidekick only, Plain meters them by effort rather than by message, and there is no published overage rate. Foundation’s 2,000 and Horizon’s 15,000 are not comparable to any competitor’s unit, so they cannot be benchmarked — only measured.

Best for

Support and CS leaders at API-first B2B SaaS companies running 3 to 10 support seats, where an engineer or CTO sits on the buying team and the support surface is already part of the product. The fit is sharpest when you want your own model answering customers against your own documentation, and you have decided not to hand deflection economics to a vendor who also defines what counts as a resolution.

Do not buy it if

Your support org is past roughly 15 people and organized around tiers, macros, queues and workforce management. Self-serve caps at 10 seats with everything beyond it behind Frontier, and the operational furniture a 40-agent floor needs — shift scheduling, calibrated QA, IVR — is not what Plain built.

Skip it if nobody on the team writes code. The API is the reason to pick Plain; a team that will only ever use the shipped UI is buying a more expensive Front.

And skip it if you run high-volume anonymous consumer support. Plain’s model is named accounts on named channels, and an unidentified end user has nowhere to sit in it.

Versus the alternatives

• Zendesk — the volume incumbent and the default when you have no specific reason to deviate. Deepest workforce tooling, largest app ecosystem, and the only option here with a credible answer for a 200-agent floor. Pick it when support headcount rather than support surface area is your scaling problem. Model its $1.50-$2.00 per-resolution AI line before comparing seat prices, because that is where the two bills actually diverge.
• Intercom — the other share leader, and the better pick when support is in-product messaging to a large self-serve user base. Fin is the strongest packaged deflection agent in the category and Intercom prices it per resolution. Pick Intercom when you want deflection working on day one and you are content not to own the model.
• Pylon — the fastest-growing entrant in B2B support and Plain’s closest pole. Same shared-Slack-channel thesis, but Pylon ships a finished application with account intelligence, churn signals and CRM sync already assembled, and has moved 150-plus teams off legacy helpdesks. Pick Pylon when you want the shared-channel product to work without engineering time. Pick Plain when you have the engineering time and want primitives instead of an opinion.
• Front — pick it when the unit of work is a conversation in a shared inbox rather than a customer account, and when the team is non-technical. Lower entry price, much less to build against.

Watch-outs

• “Ari is currently free” is a repricing waiting to happen. Free customer-facing deflection is the strongest economic argument for Plain, and the vendor describes it as a current state, not a contract term. Every competitor meters this. Guard: get the Ari rate written into the order form as a fixed price — including zero — for the full contract term, with notice required before any change. If sales will not commit to that, re-run your business case at the competitive rate of roughly $1.50 per resolution and check whether it still clears.
• The 5-seat cliff prices your sixth hire at $421/month. Foundation stops at 5 seats and the next tier opens at $299 with three seats bundled, so one additional person more than triples the bill. Guard: if you expect to cross six support seats inside the contract year, negotiate Horizon from the start and use the projected headcount to argue the discount, rather than being repriced mid-year with nothing to trade.
• Sidekick credits are metered by effort with no published overage rate. You cannot forecast the second half of the bill, and 15,000 credits cannot be benchmarked against any competitor’s unit. Guard: run the 7-day trial against your real inbox, record credits consumed per resolved thread, multiply by monthly volume, and make the vendor commit an overage rate in writing before signing. An unpriced meter is an uncapped one.
• Composability is a permanent staffing commitment. The reason to buy Plain is that you will build against the API, which means a named engineer owns the support surface indefinitely — and that engineer will eventually leave. Guard: register the Plain integration in your normal service catalog with an owning team and a runbook, and require that everything built against it be reproducible from the repository alone. If you cannot name the owner at purchase time, buy Pylon instead.
• The reference base is narrow and skews to developer tools. The published logo list is almost entirely dev-tool companies, which tells you the product fits that shape and tells you nothing about a fintech or healthcare support org carrying compliance obligations. Guard: ask for two references outside developer tooling at your seat count, and verify SOC 2 scope and data-residency options directly — the documented API endpoint is UK-hosted, which is a question to ask rather than a problem in itself.