A brand monitoring MCP server for your own agent
Nephia's MCP server lets Claude, Cursor, ChatGPT or any MCP client read the mentions your brand gets on 14 sources.
Add https://api.nephia.cc/mcp, sign in once in the browser, and ask the way you would ask a colleague: which complaints came in since Friday, what Reddit said this week.
It is on every plan, Free included, and reading your own mentions through it costs nothing.
Install it in one line
The server is remote and signs you in with OAuth. The first call opens your browser, a consent screen lists what the agent is asking for, and you decide. There is no key to copy and nothing to paste.
claude mcp add --transport http nephia https://api.nephia.cc/mcp
A question, the call it becomes
The agent turns a plain question into a call over your own mentions, with the filters the Feed already has: source, intent, sentiment, a date. The Source is an argument, not a tool, so one call reads every source you watch.
> Which complaints came in since Friday? list_mentions · intent=complaint · since=2026-09-04 - hosted checkout keeps timing out since this morning, anyone else on acme seeing this? (1,204 views) - Webhook signature check fails on Node 22 (#412) (23 reactions)
What your agent can call
| Tool | What it does | Cost |
|---|---|---|
| list_mentions | List mentions | Free in text mode · 1 credit/semantic question |
| mentions_stats | Count mentions | Free |
| mentions_similar | Find similar mentions | 1 credit/mention asked about |
| list_keywords | List keywords | Free |
| get_keyword | Get a keyword | Free |
| get_keyword_results | List a keyword's mentions | Free in text mode · 1 credit/semantic question |
| get_keyword_events | Replay a keyword's raw events or activity | Free |
| list_keyword_buckets | List a keyword's buckets | Free |
| keyword_manage | Pause, resume, or retire a keyword | Free |
| keyword_estimate | Estimate a keyword or a change | Free |
| keyword_create | Create a keyword | Metered per check |
| keyword_update | Change keyword settings | Metered per check |
| keyword_source_set | Add or replace one of a keyword's sources | Metered per check |
| keyword_source_manage | Pause or resume one of a keyword's sources | Free |
| keyword_bucket_manage | Create or delete a bucket | Free |
| keyword_brand_set | See or set a keyword's brand | Free |
| explore_estimate | Estimate an Explore run | Free |
| explore | Explore the Sources once | 1 to 65 credits/search |
| analyses_run | Run an AI pass over items | 1/50 items (agent: 1/25) |
| analyses_estimate | Price an AI pass | Free |
| vinted_price_history | Get a Vinted item price history | 3 credits |
| vinted_market_stats | Get Vinted market price statistics | 5 credits |
| account_credits | Get credit balance | Free |
Read from the server’s own registry when this page is built. Every tool description carries its cost, so a model can budget before it spends.
What the consent screen asks for
- Read your mentions and keywords
- Every free read: mentions, keywords, buckets and your credit balance.
- Change what is running
- Create, edit, pause, resume or retire a keyword and its sources.
- Spend your credits
- Search by meaning and AI passes. It arrives unticked: an agent without it still reads everything, and says so when a call needs it.
- Disconnect
- Settings, API keys, Connected agents. The agent has to sign in again to reach your account.
The mentions behind that answer
From the demo account. Acme and every handle are fictional.
X@claire_dev1,204 views
hosted checkout keeps timing out since this morning, anyone else on acme seeing this?
negativecomplaint
Names a checkout outage and asks whether others see it, a complaint with reach, not a one-off.
GitHubacmedev/acme-node23 reactions
Webhook signature check fails on Node 22 (#412)
negativecomplaint
A reproducible failure against the SDK with a runtime version, a bug report, filed as a complaint.
Questions
Do I need an API key?
No. The remote server signs you in with OAuth: your browser opens once, you choose what the agent may do, and the connection is live. Spending credits is a separate permission, left off unless you tick it.
Which clients does it work with?
Any client that can add a remote MCP server. In Claude Code it is one command; in Cursor it is the URL in mcp.json; in Claude.ai and ChatGPT it is a custom connector pointed at https://api.nephia.cc/mcp.
Can my agent set up monitoring?
Yes. `keyword_create` builds a whole keyword from the agent: its terms, its Sources, its interval and an optional webhook you control, after showing you the estimate. It also reads every keyword and can pause, resume or retire one, which is how it stops one spending. Rules, buckets and delivery channels stay in the dashboard.