Hi HN. I run a one-person company in Taiwan. MindThread is a Threads scheduler with paying users here, and it already had a REST API. This release wraps it for agents: a CLI (`npx mindthread-cli`) and an MCP server (`npx -y mindthread-mcp`, or hosted at https://mindthread.tw/api/mcp, registry id io.github.ppcvote/mindthread).
Your agent can draft posts in the account's voice, publish or schedule text, images, carousels and video, work a queue, read insights, and review reply drafts from "cloud patrol", which searches Threads for your keywords. Everything in this release goes through Meta's official Threads API, so no passwords. (A separate, Taiwan-only browser extension exists and is not part of this.)
The part I care about most is control. Each key has allow, ask or deny per permission (publish now, schedule, send replies, turn on patrol). With ask, the call returns `approval_required` and a link, and nothing goes out until the owner confirms on mindthread.tw. In the CLI that is exit code 3, so the agent stops instead of retrying. Defaults for a new live key: publish now, sending replies, enabling patrol and loosening guardrails ask first; scheduling a post 30+ minutes ahead is allowed, and you can set it to ask. Content rules (length, banned phrases, only numbers from your fact list) are checked by the server on every API write and again at publish time. With `enforce: block` a violating post is refused; the default is `warn`, which only flags it.
Limits: Threads only. Free plan is 1 account, 3 posts a day and 50 calls a day per live key, and Meta's per-profile caps apply on top. The MCP server lacks the thread, webhook and A/B tools the CLI has. The CLI and bridge are MIT; the service is hosted.
No signup needed to look around: `curl https://mindthread.tw/api/mcp` lists the tools, and `tools/list`, `get_playbook` and `list_formulas` work without a key. A free account gives you a `mt_test_` sandbox key.
A `mt_test_` key simulates every write. Feedback I want: is per-permission ask the right granularity, and what would you need before letting an agent post for you?
Hi HN. I run a one-person company in Taiwan. MindThread is a Threads scheduler with paying users here, and it already had a REST API. This release wraps it for agents: a CLI (`npx mindthread-cli`) and an MCP server (`npx -y mindthread-mcp`, or hosted at https://mindthread.tw/api/mcp, registry id io.github.ppcvote/mindthread).
Your agent can draft posts in the account's voice, publish or schedule text, images, carousels and video, work a queue, read insights, and review reply drafts from "cloud patrol", which searches Threads for your keywords. Everything in this release goes through Meta's official Threads API, so no passwords. (A separate, Taiwan-only browser extension exists and is not part of this.)
The part I care about most is control. Each key has allow, ask or deny per permission (publish now, schedule, send replies, turn on patrol). With ask, the call returns `approval_required` and a link, and nothing goes out until the owner confirms on mindthread.tw. In the CLI that is exit code 3, so the agent stops instead of retrying. Defaults for a new live key: publish now, sending replies, enabling patrol and loosening guardrails ask first; scheduling a post 30+ minutes ahead is allowed, and you can set it to ask. Content rules (length, banned phrases, only numbers from your fact list) are checked by the server on every API write and again at publish time. With `enforce: block` a violating post is refused; the default is `warn`, which only flags it.
Limits: Threads only. Free plan is 1 account, 3 posts a day and 50 calls a day per live key, and Meta's per-profile caps apply on top. The MCP server lacks the thread, webhook and A/B tools the CLI has. The CLI and bridge are MIT; the service is hosted.
No signup needed to look around: `curl https://mindthread.tw/api/mcp` lists the tools, and `tools/list`, `get_playbook` and `list_formulas` work without a key. A free account gives you a `mt_test_` sandbox key.
A `mt_test_` key simulates every write. Feedback I want: is per-permission ask the right granularity, and what would you need before letting an agent post for you?