Customer.io vs Dreamlit

Customer.io 45

No x402, MPP or L402 (0). Public plan prices with per-unit overage, Essentials from $100 a month, $0.009 per additional profile and $0.12 per additional 1,000 emails. Premium starts at $1,000 and Enterprise is priced by sales, so 15 of 20. A free 14-day trial, per the plan features page. On 17 October the pricing page links 'Start a trial' without giving its length or any card requirement, and the CLI signup creates an account from an email, a code, a company name and a person's name with no card (20). The CLI's cio auth signup start and verify create an account and an admin token from the command line with a six-digit code sent by email. That needs an inbox but no browser. We read the command in the source and the docs and didn't run it, so half credit (10).

Dreamlit 10

No x402, MPP or L402 (0). Plan prices with included emails are public in pricing.md, no per-email overage price (10 of 20). No free plan for new accounts (0). Signup, database connection and payment happen in the web app (0). Total 10.

Customer.io 86

Public OpenAPI 3.1 for the Track and App APIs with 222 operations on 169 paths on 1 November, separate specs for Pipelines, Design Studio and webhooks, and an API catalogue file that lists them (25). llms.txt, per-section indexes, and Markdown for any page by adding .md (10). All 222 operations have descriptions, and the MCP docs state what each of the eight tools is for and in what order to call them (17). 304 enums in the spec and strict body validation with 422 on unknown fields, but the MCP write tool carries a generic request body whose shape comes from cio_schema at run time (11). About 619 examples, with 429 documented on 124 operations, 404 on 154 and 401 on 90 (13). Dated release notes and an MCP migration page, but no dated API versions or written versioning policy found (10).

Dreamlit 59

The registry entry and the dreamlit-ai/dreamlit-mcp repository carry metadata only, so tool JSON Schemas are readable only inside a session; the docs give each tool's purpose and required scope (5 of 25). llms.txt, pricing.md and Markdown copies of docs pages (10). Tool descriptions say when to use each, such as edit_workflow_steps for exact edits without AI, and the publish pair's order (16 of 20). Parameter types aren't published beyond names such as includeEmailHtml and startNewChat (5 of 15). Example prompts and a troubleshooting section for each documented error state (10 of 15). A dated changelog; server version 0.1.0 in the registry (13 of 15). Total 59.

Customer.io 75

Graded as a hosted service. Statuspage at status.customerio.com with nine components, re-read on 1 November, and incident history (20). The history page lists six incidents in the 90 days to 1 November 2026, all marked minor or none. Data processing delays for a small subset of US customers on 3 August (07:08 to 07:42 UTC), the interface on 5 August (26 minutes, none), the in-app Agent on 3 September (5 minutes), reduced delivery to Microsoft-hosted addresses on 4 September (Microsoft's incident, none), US sending delays on 19 September (27 minutes) and the workspace interface on 27 October (19:11 to 19:36 UTC, with processing and sending unaffected). The two July incidents marked major (20 and 31 July) left the window on 18 and 29 October, and with them the 5 points taken off for frequency, so minor only (20). Limits published with numbers, 10 requests a second on the App API, one every 10 seconds for API-triggered broadcasts, and a guideline of 3,000 requests per 3 seconds on Track and Pipelines that returns no 429 (15). 429 with Retry-After on the App API and automatic backoff in the CLI, but no idempotency key found for sends or other writes (10). No SLA found, and the terms promise commercially reasonable efforts (0). The APIs are generally available and the MCP docs carry no beta label. The CLI is at v0.0.36 (10). 20+20+15+10+0+10 = 75.

Dreamlit 51

Graded with the hosted lines. A Better Stack status page at status.dreamlit.ai with 90 days of history for the Web App and the Workflow Engine, but no component for the MCP server (16 of 20). The record shows 100% for the web app and 99.992% for the workflow engine over 90 days, about ten minutes, with no written incidents (25 of 30). No rate limits found in the MCP or pricing docs (0). No 429, retry or idempotency guidance found (0). Scale includes dedicated support with an SLA and Enterprise has custom SLAs, but no published availability commitment (0). The MCP server, introduced on 5 May 2026, carries no beta label (10). Total 51.

Customer.io 76

Closed service with published terms, revised August 2026 (re-read 1 November). The CLI source is public under Apache 2.0 with the Commons Clause, which isn't an OSI licence, and the Claude Code plugin is MIT (15). Privacy policy revised August 2026 and a DPA effective September 2026 that deletes company data at termination on request. The AI settings page says customer data isn't used to train models. Retention is stated as the length of the agreement, with no numbered periods (22). The MCP migration page documents the May 2026 tool changes and a 30-day window on request. The 30 October key change says existing credentials keep working, with no end date. No general API deprecation policy found (8). Sub-processor list revised October 2026 with purposes and locations, US and EU regions, and 10 days' notice of changes in the DPA (20).

Dreamlit 59

A closed service under terms that name Notika AI (d/b/a Dreamlit AI) (15 of 30). The privacy policy keeps workflow execution history and delivery events for up to a year, while the security docs say Dreamlit retains no copy of customer data and cleans up after processing, which don't agree (15 of 30). No deprecation policy or dated notices found (0). The privacy policy points to a current subprocessor list and SCCs in a DPA, and names the United States and other countries as locations (12 of 20). Total 42.

Customer.io 83

Graded on the MCP server. Eight tools, with endpoint schemas fetched on demand through cio_schema in place of one tool per endpoint. cio_prime loads a reference at the start of a task, which costs some context (23). Filtering, pagination and auto-pagination on cio_read_api, and start and limit cursors on the App API (18). Tool errors are structured JSON, unknown fields return 422, and the CLI has stable exit codes (17). Reads, writes and deletes are separate tools and writes and deletes support dry-run previews. No idempotency keys, and we couldn't read the tool annotations without an account (10). Official libraries for Node, Python, Go and Ruby (15).

Dreamlit 67

16 tools, the 11 to 30 band; tools follow the granted scopes, so a read-only grant sees fewer (20 of 25). Analytics take filters, sorting and pagination, and brand styles paginate (18 of 20). The docs describe the errors an agent will meet, such as authorization required, choose a project and browser hand-off, without codes (12 of 20). Publishing is prepare then confirm, and tests don't publish; no idempotency or annotations documented (10 of 20). Empty scope requests default to the v1 scopes and get_status gives a first call; no SDK or REST route (7 of 15). Total 67.

Customer.io 79

OAuth with dynamic client registration, PKCE, five scopes and a revocation endpoint, per-user sessions that can't exceed the user's role, and service account tokens with expiry, a permanent read-only option and exchange for one-hour JWTs. On 1 November the authorisation server metadata lists only the S256 PKCE method, where on 17 October it also listed plain, so the 2 points taken off for it come back. The public and private keys introduced on 30 October split browser keys (pk_, no sends or reads) from server keys (sk_), and keys are rotated by adding a new one and revoking the old. A private key has no per-key scopes and works with every API, sends and deletes included, so as with grist, cal-com and calendly (scoped OAuth beside a full-access key) we take 2 off. Service account tokens can be read-only or limited by role, and expire (28). Connections default to read. Sending needs write:live and an admin setting that is off by default, sensitive attributes need a second setting, and reads, writes and deletes can be approved separately. No server-side approval step once live writes are allowed (18). Profile attributes and events are customer-written. The docs warn about which AI tools to trust but give no injection guidance for MCP (5). Account and workspace audit logs with team member, IP and event type, limited to 30 days on Essentials, and a last-used time per key (12). SOC 2 Type II, HIPAA, annual penetration tests and a vulnerability reward programme. No security.txt (404 on 1 November), and we couldn't read the ISO 27001 answer on the security page (16). 28+18+5+12+16 = 79.

Dreamlit 74

OAuth to one workspace with four scopes, and project-scoped personal access tokens with chosen scopes, shown once (30). A read-only grant is possible, publishing needs its own scope, confirm_publish runs only after the user approves the prepared payload, and a dedicated least-privilege database user reads the customer's data (20). The docs warn that AI clients can be tricked by untrusted prompts and tell users to review edit, publish and unpublish calls (8 of 15). The security page lists access logs for database operations and for the platform, and analytics expose run rows; no customer-facing log of MCP calls was described (8 of 15). The security page claims penetration testing and a bug bounty programme without a link, SOC 2 is coming soon, no security.txt (404), and the MCP repository's SECURITY.md routes reports to support (8 of 20). Total 74.