Run your AI-search crew from the assistant you already use

Attensira has an MCP server at mcp.attensira.com and a REST API at api.attensira.com. Connect Claude, ChatGPT, Cursor, VS Code or n8n, and your assistant can set up a workspace, read every number with its sample size, approve the crew's work and hand it more. Reading is free; only tools that start agent work spend credits.

Start free Book a demo

7 days free · card required

Your assistantattensira · attensira.com

Asked: What has the crew made for us that is not live yet?

called list_deliverables

Ten pieces, every one waiting on us to put it live. With no publishing connection, each is ready to paste into the site.

New article: How to measure share of voice in AI answers: 3 Oct; Article · ready to paste
Rewrite a passage on Best Profound Alternatives: 4 Oct; Page fix · ready to paste
New article: Does my website show up in Perplexity?: 30 Sep; Article · ready to paste

Our own workspace, three of the ten newest. Nothing ships until a person puts it live.

Writer · Wrote themattensira.com workspace, read over the Attensira MCP · 4 Oct 2026

Who does the work

Writer · Wrote the articles waiting to go live
Copywriter · Rewrote the passages
Chief of staff · Hands each line you send to a project

What can your assistant do with it?

Everything you would open the dashboard for, without leaving the conversation you are in.

Set up a business
Give it your domain. The agents draft the profile, the rivals, the platforms and the goals from your site, and you keep what is right. confirm_setup returns a checkout link: payment always finishes in your own browser.
start_workspace · confirm_setup
Read every number
Mention and citation rates, sources, competitors and pages, each with the n it was measured from, and the verbatim answers behind them. A change that is not statistically real comes back marked real: false.
get_analytics · get_prompt · deciding_sources · answer_history
Approve and steer
See what is waiting on you, approve or decline it, steer a project in plain words and list what shipped. Anything your assistant files or approves is marked with its name in the app, such as via Claude.
list_inbox · approve_work_item · steer_project · list_deliverables
Hand the crew work
Chief of staff
Send work as lines, one piece per line. The Chief of staff hands each line to the project whose crew does it. Tools that start agent work spend credits, these three among them; “What does it cost to use?” below lists every one.
ask · run_automation · add_prompts

What does one read look like?

One call, get_analytics area=competitor, returns every rival named in the same answers as you, each as a count over the answers read, and the window it covers. Reading it is free.

get_analytics · competitorsn=4,518 answers

ChatGPT and Perplexity, 61 prompts · 4 Sep 2026 to 4 Oct 2026
Name	get_analytics · competitors	Count
Semrush	35.9%	1,621
Profound	31.4%	1,417
Ahrefs	25.9%	1,171
Peec AI	21.8%	985
Otterly AI	21.3%	964
attensira.comyou	4.4%	197

attensira.com workspace, read over the Attensira MCP · 4 Oct 2026 · n=4,518 answers

How do I connect it?

One URL for every client. What changes is where you paste it, and whether you sign in or use a key.

Claude Code, one command

claude mcp add --transport http attensira https://mcp.attensira.com/mcp

Step 1
Add the server
Point your client at https://mcp.attensira.com/mcp. It speaks Streamable HTTP, which current Zed and the other clients here add as a remote server; only a client that can launch local processes alone needs the mcp-remote bridge.
Step 2
Sign in, or use a key
Where a browser can open, sign in with OAuth and nothing secret lands in a config file. For n8n, CI and scripts, mint a workspace API key under Settings → Your assistant.
Step 3
Check it with get_account
Ask your assistant to call get_account. It lists the workspaces the connection reaches, the credit balance, the scopes and the platforms that are really measured.

Step by step, per client

What is waiting on you?

list_inbox returns what the crew has parked for a person, and approve_work_item or decline_work_item answers it from the chat. These three wait in our own Inbox; none ships until someone says yes.

Inbox3 waiting for you

ApproveNew article: AI crawler logs analysis tool: what to use, and what the log actually tells you
Waiting since 2 Oct
Choices:Approve,Decline
ApproveNew article: AI prompt volume tool
Waiting since 1 Oct
ApproveRewrite a passage on Attensira — Whatever it takes to win the answer.
Waiting since 30 Sep

attensira.com Inbox (role shown is the one whose tool makes the item), read over the Attensira MCP · 4 Oct 2026

What does it cost to use?

Reading is free, however often you call it. Credits meter agent work only (1 credit = 1¢).

See plans and credits

Free: Every read: analytics, prompts, pages, sources, win plans, the answers themselves, the Inbox, deliverables and goals. Creating an automation is free too.
Spends credits: ask, run_automation and add_prompts; add_context with run: true; answer_inbox_question when the run that asked has finished; and the crew's reply to a steer_project post.
Never from chat: No assistant, API key or MCP connection can set or raise your monthly budget, and no plan is charged from a conversation. Checkout finishes in your browser.
Which plans: Every plan: the MCP server is not a tier feature. Three plans (Starter, Growth and Business) plus Enterprise differ mainly in how many prompts you optimize towards (50, 150 or 350) and how many credits each includes (300, 600, 1,200 a month). We sell credits, and you can buy more any time.

What keeps an assistant inside the lines?

A connected assistant is a workspace credential, not a read token. These are the rules it works under.

OAuth carries read and write: There is no read-only consent. Restrict a signed-in assistant by what you ask it and by your client's tool allow-list. For a credential that cannot write, mint a read-only API key.
Deleting is named: Only remove_prompt, remove_competitor, delete_automation and delete_session delete anything. They are marked destructive, so a client can ask you before it calls one.
Your organization only: One connection reaches every workspace your organization owns. A workspace id from anyone else's comes back as one that does not exist.
Approval still applies: The crew's work follows the workspace's approval mode, Draft by default. Anything stating a claim, such as a price or a competitor's name, waits for a person in every mode.

n8nmcp.attensira.com · read-only key

Asked: Add “best invoicing app for freelancers” to our tracked prompts.

called add_prompts · read-only API key

Refused. A read-only key reads everything and writes nothing, and its scope cannot be widened later.

Response: 403 forbidden
Challenge: insufficient_scope; The key is valid; it may not write. A 403 is not a reason to drop it.
To write: a new key; Mint a read and write key, or sign in with OAuth.

An example of the rule as docs.attensira.com/mcp/tools states it; the prompt is made up.

Example · the scope rule as docs.attensira.com/mcp/tools states it

MCP or the REST API?

Twins on purpose: every REST endpoint has an MCP tool with the same payload, and both take the same atn_live_ key.

MCP server

An assistant is doing the work: Claude, Claude Code, Cursor, ChatGPT. It discovers the tools and decides which to call. OAuth or a key.

REST API

Your code is doing the work: a nightly export, a deploy check, an internal dashboard, an n8n HTTP step. A key only, at api.attensira.com/v1.

Questions about the MCP server

Does Attensira have an API or MCP server?

Yes. Attensira has an MCP server at mcp.attensira.com/mcp and a REST API at api.attensira.com, so you can run the crew and read every number from Claude, ChatGPT, Cursor or your own code. Every REST endpoint has an MCP tool with the same payload.

Can my assistant set up Attensira for me?

Yes. Give it your domain: start_workspace drafts the profile, rivals, platforms, goals and plans from your site, and you keep what is right. confirm_setup saves your choices and returns a checkout link. Nothing is charged by the assistant; payment finishes in your browser.

Which Attensira MCP tools cost credits?

Reading is free, however often you call it. ask, run_automation and add_prompts spend credits, as do add_context with run: true, answer_inbox_question when the run that asked has finished, and the reply to a steer_project post. Creating an automation is free; each run it makes spends.

Should I sign in with OAuth or use an API key?

Sign in wherever a browser can open: Claude, Claude Code, Cursor, VS Code and ChatGPT. There is no key to leak into a config file, and removing the server revokes it. Use a static atn_live_ key for headless callers such as n8n, CI jobs and scheduled scripts. Both work on the same endpoint, permanently.

Can I give an assistant read-only access?

Not over OAuth: a signed-in connection always carries read and write. Mint a read-only API key under Settings → Your assistant instead. A write tool called with it returns 403 insufficient_scope, and a key's scope cannot be widened later.

Can one connection reach several workspaces?

Yes. The credential names your organization, and every workspace tool takes an optional project, so a customer running several brands connects once. get_account lists the workspace ids. A workspace outside your organization looks exactly like one that does not exist.

Why does my client show no Attensira tools?

The server speaks Streamable HTTP only, so a stdio-only client needs the mcp-remote bridge. Some clients load servers only into new conversations, or cache the tool list until a restart. https://mcp.attensira.com/healthz answers without a key if you want to confirm the server is reachable first.

Run your AI-search crew from the assistant you already use

What can your assistant do with it?

Set up a business

Read every number

Approve and steer

Hand the crew work

What does one read look like?

How do I connect it?

Add the server

Sign in, or use a key

Check it with get_account