Threads posts with DeepSeek and Bolta
Use a DeepSeek model as the drafting brain, with Bolta holding the approval gate. Parts of the DeepSeek route are not confirmed yet, and the note below says which. Nothing goes live on Threads until you say yes.
Partly confirmed, see what is and is notChecked 2026-11-21
DeepSeek is the model. The Claude Code guide shows how to connect Bolta to the agent that runs it.
Draft a Threads post about what we shipped this week, under 500 characters, and show me the text first
Illustration of the reply: Bolta returns a draft showing the connected account and the scheduled time, with Approve, Edit and Reject actions.
What is and is not confirmed
Read this before you set up DeepSeek
- Confirmed from DeepSeek's own docs: its API has an Anthropic-compatible endpoint at https://api.deepseek.com/anthropic, and a guide for pointing Claude Code at it with your DeepSeek API key.
- Confirmed from DeepSeek's own docs: that endpoint ignores mcp_servers, so DeepSeek itself never holds the Bolta connection. The host agent does.
- Not found in DeepSeek's docs: any connector or MCP feature in the DeepSeek chat app, or a remote MCP setting in DeepSeek Harness.
- Not tested by us: how well Bolta's tools behave when the host agent runs a DeepSeek model. Draft quality and tool-call reliability can differ from other models, so review every draft.
What Bolta does on Threads
Threads caps a post at 500 characters, and Bolta handles longer text by splitting it into a thread of connected posts, so one long idea does not fail.
- Text posts, and posts with images.
- A carousel of 2 to 10 items.
Threads limits and format rules
Network rules last checked 2026-07-28. Bolta's posting rules for your workspace (get-posting-rules) win if they differ from this page.
Not promised for Threads from DeepSeek: Polls on Threads. Not confirmed for Bolta. Mirroring a Threads post onto Instagram. Threads does not do that on its own, and this page does not confirm that Bolta does it.
| What to know | Threads's own rule | What Bolta does |
|---|---|---|
| Post length | A single Threads post is capped at 500 characters, and a reply carries the same limit. | Bolta uses 500. Longer text is split into a thread of connected posts. |
| Media | A post can carry up to 20 photos or videos in one carousel. Video is limited to 5 minutes and 1 GB. | A carousel in Bolta takes 2 to 10 items. An image over 8 MB is shrunk if Bolta can. |
| Links | Links are allowed, show as a preview card, and are capped at 5 per post. | Each Threads link is swapped at publish time for a short Bolta link that counts clicks. If the swap fails, the link goes out as written. |
| Continuing past the limit | You can continue by replying to your own post, which keeps the follow-on in the same chain. | Send a long text as several parts and Bolta posts every part after the first as a reply, which keeps them in one chain. |
| Formatting | Not stated on this page. | Plain text only on Threads. Markdown is flattened, and tags and mentions stay as typed. |
DeepSeek jobs for Threads, and how to write them
- Set up a recurring routine from a plan you read and confirm, instead of clicking through settings
- Write one longer text when you want a thread, and read where it splits before you say yes.
Example prompts for Threads
Things to type in DeepSeek. Each one asks for a draft and shows you the text first.
Draft a Threads post about what we shipped this week, under 500 characters, and show me the text first
What is waiting for my approval in Bolta right now?
Write one longer text about our founding story that Bolta can split into a thread, and show it to me before saving
List my Bolta brands and the accounts connected to each one. Do not change anything.
Bolta saves a draft, you read the exact text, you say yes
- 01
Bolta saves a draft
DeepSeek asks Bolta for the Threads post in your brand voice (save-post). A saved draft is never public.
- 02
You read the exact text
On Threads the exact text includes where a long text is split into a thread and the topic tag, so read each part in order.
- 03
You say yes
Only then does Bolta schedule or publish (publish-post), in your workspace's timezone. Safe Mode, client sign-off and brand approval rules still apply.
Using DeepSeek for Threads
Parts of this route are not confirmed. The note above says which.
For Threads, DeepSeek, from DeepSeek, reaches Bolta through Through a host agent running a DeepSeek model (Claude Code is the best documented route). Sign-in: Sign in (OAuth) in the host agent, or a Bolta API key. The Threads account itself is connected inside Bolta, not in DeepSeek.
DeepSeek is a model, not an app you add Bolta to. The connection lives in a host agent, such as Claude Code running a DeepSeek model, so the host saves the Threads draft to Bolta and Bolta keeps the approval gate.
An illustration from the DeepSeek guide, not real output. You ask: Show me my voice profile for Acme Studio, and tell me what is missing from it. The reply goes along these lines: Here is the voice profile Bolta holds. It has tone notes but no example posts you like. I changed nothing. Want me to suggest an update for you to confirm? Read only. Edits need your yes.
Until someone tests it, we do not promise a Harness setup. Use a host agent instead.
DeepSeek's Claude Code guide maps its own models onto the model-name variables Claude Code expects. We list ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN on this page because they are stable, and leave the model-name values to the guide, since those may be renamed.
We do not know DeepSeek's retention terms, so for client work read them before you route a brand's notes through the model.
- Draft posts for a chosen brand. Ask for one post or a batch. The agent checks the posting rules for the network, writes in your voice and saves each result as a draft you can open in Bolta.
- Review and approve on your terms. List what is waiting, read each draft, and approve only the ones you name. Approval uses the content fingerprint, so a draft edited after you read it cannot slip through.
- The host agent holds the Bolta connection. Claude Code keeps the MCP link to https://mcp.bolta.ai/mcp and makes the actual tool calls. DeepSeek's endpoint ignores mcp_servers, so this layer cannot be skipped.
- It pulls your voice before writing. The agent asks Bolta for your voice context, so the DeepSeek model writes in your tone and not in a generic one.
- Setup: Pick a host agent that supports MCP. Claude Code is the best documented option for DeepSeek. DeepSeek also publishes guides for Codex, OpenCode, OpenClaw and Hermes. This page uses Claude Code.
- Setup: Get a DeepSeek API key. Create the key on the DeepSeek platform. This key pays for the model only. It is not a Bolta key and Bolta never sees it.
- MCP in DeepSeek Harness. The Harness docs we read show a plugin system but no remote MCP setting. Third-party plugins that claim MCP support are untested against Bolta.
- A routine you confirmed lives in Bolta and repeats without the agent open.
- The DeepSeek chat app or API on its own. Choose this if you hoped to add Bolta straight to DeepSeek. We found no connector system in the chat app, and DeepSeek's Anthropic-compatible API ignores mcp_servers. There is no verified way to add Bolta here.
Connect DeepSeek to Bolta
DeepSeek is the model. The Claude Code guide shows how to connect Bolta to the agent that runs it.
Threads questions, answered for DeepSeek
- What does Bolta do when DeepSeek writes more than 500 characters for Threads?
- Bolta splits longer text into a thread of connected posts, and the 500 limit also applies to each image caption. The parts are posted as replies in order, so read where the splits fall before you say yes.
- Can DeepSeek reply to someone else's Threads post?
- Yes, from a public link (reply-to-social-content). The agent writes the text first with the voice-aware draft tool (draft-engagement-reply) (for a stored opportunity) or (draft-inbound-reply) (for a comment on your own post), shows it to you, and calls reply-to-social-content only after your yes, because that call posts the reply. Finding conversations to join (search-engagement-opportunities) needs a campaign already attached to the account.
- I get an authentication error from the model
- ANTHROPIC_AUTH_TOKEN must be your DeepSeek API key, not a Bolta key or an Anthropic key. Check it on the DeepSeek platform and that the key has not been deleted.
- Bolta does not appear in /mcp
- Run the claude mcp add command again from the same project or user scope you are using, then restart Claude Code. If you also installed the plugin, you may have two entries, so keep one.
- The agent names the wrong brands
- Brands are chosen at sign-in. Reconnect through /mcp and tick the brands you want, then ask again and name the brand in the request.
- Bolta tool calls fail or come back malformed
- This is the part we have not tested. Retry once with a simple read, then try the same request on the agent's default model. If only the DeepSeek run fails, use it for drafting text and keep Bolta steps on the other model.
- Does DeepSeek publish Threads posts without my yes?
- No. Nothing is approved, scheduled or published until you say yes to that Threads post. Whether DeepSeek completes the connection is not fully confirmed, so read the note above first.
More from DeepSeek
DeepSeek can work with the other networks Bolta publishes to. Each page has its own limits and review notes.
Threads from other AI agents
The same Threads limits apply whichever agent asks. The connection and the setup are what change.
Try a Threads draft from DeepSeek
Connect DeepSeek, ask for a Threads draft and read the exact text. Nothing goes live until you say yes.