Claude Code, Claude Desktop, Cursor, or anything that speaks MCP: paste one config and your agent can plan, create, schedule, and publish your social. You keep the review. It does the work.
1. Mint a key. In the app, open Business → API & MCP and grant only the scopes the task needs. The key is shown once; Taka keeps a hash.
2. Add the server. One command in Claude Code, or paste the config block into any MCP client's settings.
3. Ask. 41 tools, guided prompts, and live resources are now in your agent's hands.
claude mcp add --transport http taka https://api.taka.ai/mcp \
--header "Authorization: Bearer taka_live_..."{
"mcpServers": {
"taka": {
"type": "http",
"url": "https://api.taka.ai/mcp",
"headers": { "Authorization": "Bearer taka_live_..." }
}
}
}“Plan my social week and schedule the drafts for my review.”
“Make a reel for the weekend special and queue it for Saturday morning.”
“What performed best last month? Draft two more like it.”
Long work comes back as a job your agent polls, so nothing blocks the conversation while a reel renders.
Guided prompts like plan_my_week walk your agent through a full weekly plan, drafted onto your calendar for your approval.
Images, carousels, and reels, written and designed on your brand. The same engine the app uses, behind the same tools.
Queue posts to their slots or publish now. Scheduling requires the publish scope, because a schedule ships the post when its slot arrives.
Read the brand, refine the tone, swap the logo. What your agent adjusts, every future post inherits.
Analytics behind analytics:read, so your agent can see what worked before it drafts what comes next.
Keys carry scopes: generate and publish are never implied. A read-only key can report all day and ship nothing.
Two things stay in the browser on purpose: connecting a social channel (OAuth) and billing. Everything else, your agent can drive.
Revision results say what actually happened: revisionApplied false means the quality floor kept the old version, noVisibleChange means the same words came back. Your agent narrates from results, never from intentions.
Generation returns a job to poll. Statics land in 1-3 minutes, reels can take 20, and stuck jobs fail themselves with a timeout code, so polling always terminates.
An Idempotency-Key on writes replays the original response for 24 hours instead of double-charging credits or double-publishing a post.
llms.txt has the whole surface in plain text, and the taka-agent skill gives Claude Code and Cursor full instructions in one install.
Start free, mint a key, and hand your social to the assistant you already trust.