SocialForge: social media content API

A social media content API lets a program create posts for social networks. Your application sends an idea, an article or a topic. The API returns finished text, images and video in the format each network expects.

Two jobs often get confused. A publishing API takes a post you already made and sends it out. A content API creates the post in the first place. SocialForge does the creating and adds publishing through a connected service, so one call can cover the whole path.

This suits a product that needs social content as a feature, an agency serving many brands, or a team that wants posts generated from its own articles.

A call names the content and the targets.

{
  "content": "Why most small businesses overpay for insurance.",
  "targets": [
    { "platform": "instagram", "formats": ["feed", "carousel", "reel"] },
    { "platform": "tiktok", "formats": ["video"] },
    { "platform": "linkedin", "formats": ["text", "document"] },
    { "platform": "x", "formats": ["thread"] }
  ],
  "language": "en",
  "webhookUrl": "https://yourapp.example/webhooks/socialforge",
  "maxCostUsd": 10
}

The API validates the targets, estimates the cost and answers at once with a 202 status.

{
  "generationId": "66a9...",
  "status": "processing",
  "targets": 7,
  "costEstimate": { "total": 8.04 },
  "statusUrl": "/api/apps/generation/66a9.../status"
}

Generation runs in the background. Your application polls the status URL or waits for the webhook. Targets run in parallel, four at a time.

A tool that resizes one post for each network produces the same message everywhere. SocialForge plans the campaign first, so each platform carries its own part of the story. That plan stays with the generation, so a post regenerated later keeps its place.

The content matrix is the source of truth for 13 platforms and 44 format pairs. GET /api/v1/capabilities returns all of it, so a client can build its interface from live data.

Each format entry states its text limit, media rules, aspect ratios, required fields and default models. A wrong combination fails at validation, before any money is spent.

You can leave model choice to the defaults, or set it per format, per platform or for the whole request. The catalogue holds 18 image models and 16 language models.

The API puts a price on a campaign before it runs. You get an estimate per target, and you can stop there. Video usually takes most of the bill.

maxCostUsd sets a hard ceiling. When an estimate runs over, targets are removed and the response lists them. Each finished generation carries a cost audit with the estimate, the actual cost and what was dropped.

You can redo one weak reel and leave the rest of the campaign as it is. One call reruns a single target.

POST /api/v1/generations/:id/regenerate/instagram.reel

{ "models": { "video": "veo-3.1" }, "durationSec": 25 }

The stored strategy stays in place, so the post keeps its role in the arc. The response returns the new result and the cost difference.

SocialForge publishes through a third party publishing service, once a client has connected its social accounts.

Most teams should keep a person in the review step, especially on regulated subjects. Generated text can contain errors, and a brand carries the responsibility for what it posts.

Analytics from the publishing service can be stored against each post after 24 hours and 7 days. Your application then has data to compare what worked.

Each client authenticates with an API key in the X-API-Key header. External social routes allow 60 requests a minute per client. You request a key by sending a brief.

Polling works for a small tool. A product should use webhookUrl, so results arrive when ready. A worker on your side then has a clear flow.

1. Store the generationId with your own record.
2. Show the post as processing.
3. When the webhook arrives, read each entry in outputs.
4. Save the text and media URLs, then show them for review.
5. Offer a regenerate action per target.

Each output entry holds the target key, the text and its media URLs.

An agency publishes one article a week for a client and wants four posts from it.

Without video, the three text and image targets stay cheap. The video target usually takes most of the bill, so the agency can test the three first, then add video to the articles that perform. maxCostUsd protects the account when a request grows.

The API works from what you send. A vague topic gives a vague post.

Check each figure in generated text against your source. On regulated subjects, keep a legal reviewer in the loop.

A small business posting twice a week will be better served by a scheduler. SocialForge makes sense when content generation is part of a product, or when many brands run through one system.

Brand details travel with the call, so text and images share a voice across every format.

Each platform has its own habits, so the same post rarely works everywhere.

These habits change, so read your own numbers and adjust.

1. Start from a source your team trusts: an article, a report or a customer question.
2. Generate one campaign and read it through.
3. Correct facts and tone.
4. Schedule the posts across a few days.
5. Read performance after 24 hours and after a week.
6. Feed what worked into the next brief.

Volume alone rarely helps. A few strong posts on a steady schedule tend to outperform a daily stream nobody reads.

Platform rules on AI content differ and they change, so check them before a large campaign.

Tag your links with UTM parameters, so your analytics shows which posts bring visits.

Choose a rhythm your team can review. Two or three good posts a week on one or two networks gives you enough data to learn from. Add a network once the first one works, since each adds review time.

• Every figure matches the source.
• Names, products and prices are correct.
• The tone fits your brand.
• Images show nothing misleading.
• Links work and carry their tags.
• Regulated claims have been approved.

A person with the source next to them can do this in minutes per post.