How to do it
- Draft from the editor Advocates ask their coding agent in Cursor or Claude Code to draft a post about the sample they just pushed, over MCP, without leaving the IDE.
- Check against brand memory Brand memory holds API names, casing rules, and topics the team avoids, so drafts get the details right.
- Guardrails catch the basics Missing alt text on code screenshots, over-length posts, and blocked link shorteners fail before the draft reaches a reviewer.
- Advocate approves The advocate who owns the topic approves from the inbox. Agents can never approve their own drafts.
- Learn from insights get_insights returns per-post metrics where available, and the team applies what worked to the next plan.
New sample: streaming responses from our API into a Next.js route handler in 40 lines. Handles backpressure and client disconnects. Code, a short walkthrough, and a one-click deploy in the repo. [screenshot with alt text: the route handler code in an editor]
DevRel content is technical, so the review has to be too
Developer relations lives on trust. Developers follow advocates who are precise, admit limits, and link to working code. That makes DevRel a strong fit for agent drafting and a poor fit for unreviewed automation. The agent can read a sample repo and describe it accurately. A person has to decide whether it is the right thing to promote this week.
Recommended setup
- Connect the networks your developers use. Many developer communities are active on Mastodon and Bluesky as well as X.
- Install the MCP server in the tools advocates already live in: Cursor, Claude Code, or VS Code. The MCP config generator writes the config for you.
- Fill brand memory with product names, API casing, and preferred terms. See brand memory.
Permissions by role
Give each advocate's agent its own key with draft permission. A key per agent makes the audit log readable: you can see which agent drafted what and who approved it. If one advocate runs a weekly tips series that never changes shape, promote that agent on one network through the trust ladder. Keep launch and pricing posts draft-only.
Measuring DevRel on social
Likes are a weak signal for developer content. Tag links to docs and samples with UTM parameters from the UTM builder and watch docs visits and repo activity instead. Use per-post metrics to compare formats, such as code screenshots against short threads.
Pitfalls
- Code in images without alt text. The alt text guardrail blocks it, which also helps screen reader users.
- Hype. Put words like "revolutionary" in banned terms. Developers stop reading.
- Cross-posting the same text everywhere. Mastodon and Bluesky audiences expect a slightly different register. Draft per network; the how-to thread template is a good start.