What we observed
At 2026-10-20T15:40:11.117Z, five sequential JSON-RPC tools/list requests to https://mcp.reeling.app/mcp each returned HTTP 200. Each response advertised the same 14 named tools. The measured request durations ranged from 62 to 330 milliseconds, with a median of 67 milliseconds.
These are original observations from one local development machine. The durations include the network round trip and response parsing. They are not a service-level commitment, a sustained uptime measurement or a comparison with another provider.
Reproduce the observation
The dependency-free script sends only tools/list requests. It does not provide a credential or call a publishing, generation, analytics or account-discovery tool. The JSON output records the method, timestamp, tool names, durations and a SHA-256 checksum of the script.
Clone the example repository, inspect the script, then run the command below. Your durations and advertised tool set may differ as the service changes. Keep the timestamp with the raw result when comparing observations.
git clone https://github.com/yomariano/reeling-examples.git
cd reeling-examples
node examples/observe-mcp.mjs --samples=5 --writeThe advertised tool surface
The returned names cover project/channel discovery, publishing and status, analytics, inbox operations and generation. An advertised tool is an interface, not evidence that an account can execute it. The adapter can describe platform capabilities that still depend on authorization, plan eligibility and provider access.
Public metadata discovery is useful for checking the endpoint and inspecting the contract before configuring a credential. It does not test a full client session, authenticate a user or demonstrate successful media publication.
reeling_adapt_content_ai
reeling_create_storyboard
reeling_evaluate_content_jev
reeling_generate_video
reeling_get_analytics
reeling_get_channel_options
reeling_get_inbox
reeling_get_post_status
reeling_list_channels
reeling_list_projects
reeling_publish_post
reeling_reply_to_comment
reeling_stitch_storyboard
reeling_sync_metricsKeep account evidence separate
The earlier October 15 reference check used an authorized key for two REST GET requests. It reported one project and six channels while excluding account identifiers and credentials from the saved result. That is a different observation from the public MCP metadata check described here.
Neither observation uploaded, scheduled or published content. To evaluate your own workflow, first run offline fixtures, then inspect the exact account and channel through a deliberate read. A real publishing acceptance test requires the intended media, visibility and destination checks.
How to cite this check
Cite the observation date, exact endpoint, method and five-request sample alongside the raw JSON. Report the narrow result: public tool metadata was readable at that time. Do not describe these five samples as proof of general availability, publishing reliability or customer performance.
If you run the same script later, retain its source checksum and any failures. A partial run should remain partial; successful samples should not silently replace an unavailable endpoint or parse error.