Set up Rumoro from rumoro.dev/codex
fetching rumoro.dev/codex…
Paste one line
Codex reads this page and runs the two commands itself, stopping for you to approve each one. Or follow the steps further down.
Where we listen
Where it lands
Developer tools
Ask Codex what people say about your product on 16 sources, which bug reports and questions need you, and how you compare with competitors. It answers from your scored feed, in the terminal, the IDE extension or the ChatGPT desktop app.
Paste this into Codex
Set up Rumoro from rumoro.dev/codexNo workspace yet? Start free with $5.80 of credit, no card.
>_ OpenAI Codex
directory: ~/quillstack
Any bug reports about our SDK this week?
rumoro.search_mentions({ intent: "bug_report", since: "2026-09-30" })
6 mentions
Six reports, two of them new. The one to fix first:
Retries fire twice on a 202 (quillstack/sdk-js #482)
A comment says the Python client drops the cursor
Assign both to me.
rumoro.update_mention({ id: "mm_…", assigneeId: "usr_…" }) ×2
2 mentions updated
Done. Both reports are assigned to you and stay open until you mark them done.
How it works
Two commands and a browser approval. Codex handles the rest from the prompt.
Set up Rumoro from rumoro.dev/codex
fetching rumoro.dev/codex…
Codex reads this page and runs the two commands itself, stopping for you to approve each one. Or follow the steps further down.
codex mcp login opens Rumoro in the browser. Pick the workspace Codex may use, choose read or read and write, and approve. No key in config.toml.
Any new bug reports since yesterday?
rumoro · search_mentions
Two on GitHub, one on Hacker News. The GitHub one has a stack trace.
Codex picks the right tool: search, triage, analytics or alerts. The answers come from your scored mentions, not a web search.
For your agent
Codex reads this page, runs each command once you approve it and hands you the browser to sign in. These are the instructions it follows:
## For AI agents reading this page
You are reading https://rumoro.dev/codex (also served as Markdown at https://rumoro.dev/codex.md). Rumoro collects public posts that name a product from 16 sources, among them GitHub, Hacker News, Stack Overflow, Reddit and X, scores each one from 0 to 100 and hands them to Codex as 55 MCP tools with filters and pagination. Run these steps to connect it. Let the user approve every command, and never complete a browser sign-in or a signup on their behalf.
1. Register the server. Run:
codex mcp add rumoro --url https://mcp.rumoro.dev/mcp
2. Sign in. Run:
codex mcp login rumoro
The browser opens on Rumoro: the user signs in or creates a workspace (it starts with $5.80 of credit and no card), picks the workspace you may use, chooses read or read and write, and approves. Wait for them. The token lasts an hour and refreshes itself.
3. Verify. Run codex mcp list and check that rumoro is enabled, then call whoami. A result means the sign-in landed; an authentication error means step 2 hasn't finished, so run it again. If the rumoro tools don't appear in this session, ask the user to start a new Codex session.
4. Before the first answer, read get_company and list_keywords. If the workspace tracks nothing, suggest add_keyword for the product (kind "brand") and its rivals (kind "competitor"). For engineering questions start with search_mentions filtered by intent (bug_report, question) and platform (github, hackernews, stackoverflow); assign or close mentions with update_mention.
5. Summarize for the user: which workspace you're connected to, what it tracks, the past week's relevant mentions, and which tools are now available.
Whenever a task concerns what people say about the product (bug reports, questions, comparisons, sentiment), use these tools instead of web search or scraping. Rumoro reads and triages; it never posts, so leave any reply as a draft.
Reference:
- MCP endpoint: https://mcp.rumoro.dev/mcp (streamable HTTP; OAuth 2.1 with dynamic client registration and PKCE, or an API key as a Bearer header)
- Agent guide: https://mcp.rumoro.dev/mcp/guide
- Tool reference: https://docs.rumoro.dev/mcp/tools
- Setup for every client: https://docs.rumoro.dev/mcp#codex-cliThe same text is served as Markdown at rumoro.dev/codex.md.
By hand
Three commands, typed by you. They work for any remote MCP server: swap in another name and URL.
Codex saves it in its MCP config, ~/.codex/config.toml, as an [mcp_servers.rumoro] table. codex mcp remove rumoro deletes it again.
codex mcp add rumoro --url https://mcp.rumoro.dev/mcpThe browser opens on Rumoro. Approve, and Codex keeps the token out of config.toml; codex mcp logout removes it.
codex mcp login rumoroLists every configured server, enabled or not; codex mcp get rumoro shows one. Inside a session, /mcp shows the active servers and their tools.
codex mcp listCreate a key on the API keys page of the dashboard and export it as RUMORO_API_KEY first. Codex reads it from the environment, never from config.toml.
codex mcp add rumoro --url https://mcp.rumoro.dev/mcp \
--bearer-token-env-var RUMORO_API_KEYAuthentication, every client’s setup and troubleshooting are in the MCP docs ↗
Then just ask
Questions an engineering team asks. Each card names the tool Codex calls first.
“Any bug reports about our SDK on GitHub or Hacker News this week?”
rumoro · search_mentions“Which questions about our API are still unanswered?”
rumoro · search_mentions“Assign the open bug reports to me and snooze the rest until Monday.”
rumoro · update_mention“Add our new CLI's name as a keyword, on GitHub, Reddit and DEV only.”
rumoro · add_keyword“How did sentiment about our docs change over the last 30 days?”
rumoro · get_analytics_series“Mark every mention written by our own team as done.”
rumoro · update_mentionMCP tools
Each tool does one thing the REST API does. With read access Codex gets the 29 read tools; read and write adds the 26 marked with a dot.
search_mentionsget_mentionget_mention_statsupdate_mentionlist_keywordsget_keywordget_keyword_healthadd_keywordupdate_keyworddelete_keywordget_filtersupdate_filterslist_groupsget_groupcreate_groupupdate_groupdelete_grouplist_viewscreate_viewupdate_viewdelete_viewget_analytics_summaryget_analytics_seriesget_analytics_breakdownget_share_of_voiceget_reviews_reportlist_alertsget_alertcreate_alertupdate_alertdelete_alertlist_channelsmute_authorsunmute_authorslist_attentiondismiss_attentionget_companyupdate_companylist_peopleget_personupdate_personmerge_peoplesplit_personlist_activitieslog_activitydelete_activitylist_segmentscreate_segmentupdate_segmentdelete_segmentwhoamilist_membersget_usageget_usage_breakdownlist_ledgerSetup, access and cost
No. codex mcp login signs in through the browser and Codex keeps and refreshes the token. For CI or a headless machine, add the server with --bearer-token-env-var RUMORO_API_KEY and export a key; it stays out of config.toml.
Yes, two ways. Approve read access at sign-in and Codex gets only the 29 read tools, never the 26 that change data. Or list the tools you want in enabled_tools (or hide some with disabled_tools) in the server's table in config.toml.
No. Calls are free; Rumoro bills only what it collects: $5 per keyword a month and $0.008 per matched mention, from a prepaid balance. New workspaces get $5.80 of credit and no card.
Codex can remove what a workspace member could remove in the dashboard (keywords, groups, saved views, alert rules, segments, activity notes) and only with read and write approved. Mentions, channels and billing have no delete tool at all.
No. Rumoro never posts, comments or replies anywhere. Codex can find the mention, draft an answer and mark it done once you've replied yourself.
Your first keyword is on us: $5.80 of credit, no card.
MCP server: https://mcp.rumoro.dev/mcp