No per-seat pricing. Ever.

Yes. Self-hosting is fully open source under the AGPL-3.0 license, with unlimited users and unlimited records and no credit card required. Deploy it yourself with the published Docker Compose file — your data stays on your own server the entire time.

No — Relaticle has never charged per seat. Every plan, self-hosted or hosted, is priced per workspace, so you can add as many teammates as you need without the bill changing.

Cloud Pro is $19/mo per workspace ($228 billed yearly, or $24/mo billed monthly) and includes unlimited users and records, every supported AI model from Sonnet 4.6 up to Opus 4.7, the REST API, the 30-tool MCP server, and email support. Each workspace gets a 2,000-credit monthly AI allowance; how far it goes depends on the model and how many tool calls each reply makes (see "What counts as an AI credit?" below) — as a reference point, 2,000 credits covers roughly 2,000 simple Sonnet 4.6 replies, or around 666 Opus 4.7 replies before tool calls. New workspaces start on a 14-day trial automatically, with no card required.

CRM data itself is never capped — every plan supports unlimited users, companies, people, opportunities, tasks, and notes. The only metered resource is the AI assistant: Cloud Pro's 2,000 credits a month reset each billing (or trial) period. Once they are used up, the assistant declines new chat requests until the next reset — Cloud Pro workspaces can also buy a prepaid credit top-up instead of waiting. Nothing else in the CRM is affected.

Credit cost depends on the model and how much work a reply does — it is not flat. Each message costs its model's credit multiplier (1x for Sonnet 4.6 and self-hosted models; 1.5x for GPT 5.5 and GPT 5.4; 3x for Opus 4.7), plus 0.5 credits for every tool call the assistant makes while answering — searching, creating, or updating a record — rounded up to the next whole credit with a 1-credit minimum. A simple Sonnet 4.6 reply with no tool calls costs 1 credit; an Opus 4.7 reply that touches two records costs 4. Using the REST API or the MCP server directly, outside the built-in chat, never touches your credit balance.

Every plan can use Sonnet 4.6 and any self-hosted model you connect yourself. Cloud Pro additionally unlocks Opus 4.7, GPT 5.5 and GPT 5.4 — the models with a higher credit multiplier.

Yes — a per-minute cap shared across the whole workspace, not per person: 10 messages/minute on Free, 30/minute on Cloud Pro. It exists to stop runaway usage, not to constrain normal work.

No. Self-hosting does not disable credit metering: every workspace, self-hosted or hosted, defaults to the Free plan's 300-credit monthly allowance — the same one a brand-new Cloud signup gets. Self-hosters do have direct database access, so they can raise their own workspace's plan value, but no plan removes metering entirely (even the highest built-in plan caps out at 10,000 credits/month), and changing the plan value alone doesn't reset the current period's balance — that only happens automatically once the existing period ends.

Yes. Both options run the identical open-source codebase against the same PostgreSQL schema, so neither locks you in. Companies, people, opportunities, tasks, and notes each have a built-in CSV export, and the import wizard on the other side accepts CSV — moving between a self-hosted install and the hosted plan is a standard export and re-import, not a proprietary migration.

New workspaces start a 14-day trial automatically, with no card required. If a payment method isn't added before the trial ends, hosted access pauses and you're redirected to the billing page to subscribe. Self-hosting the exact same open-source codebase is always available as a fallback.