Skip to content
Back to commentdraft

commentdraft / docs / platforms

Connecting to comment APIs on eight platforms

Eight guides, one per platform, each written from the platform's own primary sources and each saying out loud how far the checking went. Every endpoint, scope string, quota number and policy clause carries the URL it came from and the date it was read. They came out of building commentdraft, and they are here because they did not exist.

2026-08-01
Read against live docs
8
Platforms covered
226
Distinct URLs cited
1
Connectors built

The guides document how to connect. The code connects to one of them, Facebook Pages, built and tested against fakes and never run against a live Page. Nobody has made a call on any of the eight, the one with a connector included.

The published blocker is rarely the real one

Written across all eight, one pattern shows up that no single guide would have surfaced: where the received wisdom names an obstacle, the obstacle it names is either out of date or standing in front of a different one. The section worth reading on each page is not the published gate. It is the one headed "What is still unknown".

The order worth attempting them in

Numbered by how much stands between you and a first working call, not by how much you want the platform.

01

Facebook Pages

connector shipsno permission needed
In the way
Nothing to clear. An operator who owns the Page and makes their own app needs no App Review and no Business Verification.
Reply path
Documented, and Meta documents two readings of the same call
Cost
No fee. Under an hour from a standing start, by my own estimate.
Worth knowing
Start here even if Facebook is not the platform you want. It is a route where nothing is queued and nobody reads anything, so it tells you whether the rest of the tool suits you before you spend a review cycle finding out.
Read the Facebook Pages guide
02

Instagram

no permission needed
In the way
Nothing, for your own professional account at Standard Access. The comments webhook is the exception and needs Advanced Access.
Reply path
Documented
Cost
No fee. Polling instead of webhooks is the price of skipping App Review.
Worth knowing
Same Meta app, same Graph host, same token concepts. Accept polling in the first version and the whole App Review branch disappears.
Read the Instagram guide
03

Threads

no permission needed
In the way
Nothing, for the single-operator path. Anything beyond it needs App Review, which is not reachable until a Business has completed Business Verification.
Reply path
Documented, a two-step container then publish
Cost
No fee. Business Verification is company paperwork, not a build.
Worth knowing
Gate-free for a single operator, but on its own hosts with its own two-step publish and its own two expiry clocks. More build than Instagram for the same amount of paperwork, which is none.
Read the Threads guide
04

YouTube

no permission needed
In the way
A 10,000 unit daily quota, so 200 replies a day at the ceiling, 192 if a reply costs 52. An app left in Testing gets refresh tokens that expire in 7 days.
Reply path
Documented, 50 quota units a call, or 52. Google's own page carries both
Cost
No fee, and no pricing page exists. OAuth verification, which Google says "can take up to 10 days", once anyone but you uses it.
Worth knowing
The gate here is paperwork rather than code: no review and no money for your own channel, and the approval gate this tool already has is a policy requirement here rather than a product opinion. What it costs is arithmetic: 50 units a reply, or 52, against 10,000 units a day, minus whatever polling spends.
Read the YouTube guide
05

Reddit

someone has to say yes
In the way
Access is granted rather than taken. The Responsible Builder Policy requires approval before the first call, through a support ticket.
Reply path
Documented
Cost
Free at 100 queries per minute per OAuth client id. No turnaround is published for the ticket.
Worth knowing
File the ticket early, because it is the only thing on this list where the waiting can start before the building does. Read the 48 hour deletion rule first: it reshapes what a review queue is allowed to keep.
Read the Reddit guide
06

X

someone has to say yes
In the way
Prior written approval from X before replies written by software are deployed. No published turnaround, no queue position, no appeal, and no self-serve path.
Reply path
Unsettled. A programmatic reply is permitted only when the author summoned you, and nobody has established whether replying on your own post counts.
Cost
About $3 a month at 500 comments read and 250 replies published. The $100, $200 and $5,000 figures still dominating search results are for a product X stopped selling on 2026-02-06.
Worth knowing
Two questions with no published answer: whether the approval requirement covers a tool with a person approving each reply, and whether replying on your own post counts as being summoned. The second is settleable in an afternoon with a live call and it can make the first moot, so settle it before writing any connector code.
Read the X guide
07

TikTok

someone has to say yes
In the way
The developer of record must be a company with a matching domain. A solo consultant cannot register.
Reply path
Documented, on business-api.tiktok.com
Cost
No published fee. Three reviews: "three business days", "2 to 3 business days", and a form with no published turnaround.
Worth knowing
Attempt only if the developer of record is a company. The requirement lands on whoever registers the app and not on the account being moderated, so a creator with a personal account can be covered by an app a company they work with registers.
Read the TikTok guide
08

LinkedIn

someone has to say yes
In the way
A two-stage partner review, and above it the Member Data question. The Marketing API Terms prohibit exporting Member Data to a third party and define it to include a member's comment; whether a model gateway is a third party or the operator's own service provider is unsettled, and passing the review would not answer it.
Reply path
Documented
Cost
No price published on any LinkedIn property. No SLA published for either review stage.
Worth knowing
Last, because the review is long and the question above it is unanswered. Settle the Member Data reading with somebody qualified before spending a fortnight on a partner application, since a no there ends the project rather than delaying it.
Read the LinkedIn guide

The first four need no permission from anybody. The last four each need a human at the platform to say yes, and three of the four publish no idea of how long that takes.

What checking the sources turned up

Instagram

An id that resolves nothing and raises no error

GET /me?fields=id returns the app-scoped id, not the account id. It never matches the webhook payload, resolves nothing, and fails silently. The field that works is user_id.

TikTok

The published blocker is not the real one

"TikTok has no comment API" is true of developers.tiktok.com and false of business-api.tiktok.com, a different product line with two documented comment endpoints. What TikTok refuses is not the feature but the applicant: it will not onboard individual developers at all.

YouTube

The channel owner cannot read the original text

snippet.textOriginal is only returned to the comment's author, and a channel owner reading a viewer's comment is not that. A connector gets textDisplay, which Google documents as possibly differing from the original: "it may replace video links with video titles".

Reddit

Every ampersand arrives escaped

JSON bodies replace <, > and & with entities unless raw_json=1 is passed. Tom & Jerry reaches the model as Tom &amp; Jerry, and the review page escapes it again, so a person approves a reply drafted against text nobody wrote.

The last two are the same class of bug on two different platforms: the text you receive is not the text the person wrote. Neither platform mentions it at the endpoint that returns it.

Where the rest of it lives

Each guide carries its own verification section saying which claims were read twice, which rest on a single reading, and which cannot be settled without credentials. Two further pages carry what is the same everywhere: platform-policy.md maps each safety property in the tool to the clause that made it necessary, and limits.md is what the tool cannot do regardless of platform.

The repositoryHow the tool itself is built