AIsa MCP
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| AISA_API_KEY | No | Optional API key for the AIsa MCP server. If set, it is sent as `Authorization: Bearer $AISA_API_KEY` and skips the OAuth browser flow. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| logging | {} |
| prompts | {
"listChanged": true
} |
| resources | {
"subscribe": false,
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_categoriesA | The AIsa catalogue at a glance: categories, the servers in each, tool counts, and the dedicated endpoint to connect if you only need one category. Free; no key needed. (AIsa-only: tool-router has no equivalent.) Use |
| searchA | Find AIsa data operations across SEO & AI visibility, finance, social, web search & research, sales and agent mail — 950+ APIs — by describing the task. Free; no key needed. Returns tool-router's SearchResponse: retrieval_mode (plan |
endpoint | clarification), an optional plan, and candidates with
operation_id, provider, method, path, summary, required_inputs,
price, match_reasons and details_ref — plus input_schema, so a
candidate can be passed to |
| get_detailsA | Full contract of one or more operations: arguments_schema, response_schema, read_only, side_effects, availability, price, suggested_max_price_usd and known_pitfalls. Free — a quote authenticates like a call but stops before any spend.
|
| useA | Execute one AIsa operation. Billed per call to your AIsa key. Answers in tool-router's BatchCallResult shape: successful, data or error {type, status, message, retryable}. Pinned tools in tools/list can also be called directly; this is the way to call anything found through search. |
| batch_useA | Execute up to 20 operations concurrently (tool-router's batch_use). Each item answers independently; one failure never cancels the others. Billed per call to your AIsa key. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| seo | Search and AI-answer visibility: rankings, keywords, backlinks, site health, app and marketplace listings, local business data, and how a brand shows up in ChatGPT, Claude, Gemini and Perplexity answers. DataForSEO, Semrush and Ahrefs together. 316 operations; brief + how to search and call. Optional task: what you want to find out. |
| finance | Market and company data: prices, fundamentals, filings, insider trades, crypto markets, prediction markets, and a scene agent that aligns what X is saying about tickers with how they are trading. 49 operations; brief + how to search and call. Optional task: what you want to find out. |
| social | Public social data: X/Twitter users, posts, search and trends; Instagram, Reddit and Pinterest profiles and content; YouTube search. 56 operations; brief + how to search and call. Optional task: what you want to find out. |
| search | Web search, page extraction, crawling and grounded research through Tavily, Firecrawl, Exa, Perplexity, Oxylabs, txyz, and Anthropic/OpenAI grounded search. 30 operations; brief + how to search and call. Optional task: what you want to find out. |
| sales | People and company search, enrichment, sequences and CRM objects through Apollo, creator discovery, and Similarweb market intelligence: any domain's traffic, audience, rankings and search rivals. 79 operations; brief + how to search and call. Optional task: what you want to find out. |
| Inboxes, threads, drafts and sending for agents through AgentMail. 49 operations; brief + how to search and call. Optional task: what you want to find out. | |
| gtm | Go-to-market data in one set: people and company search (Apollo), creator discovery, X/Twitter, Instagram, Reddit, Pinterest and YouTube, Similarweb market intelligence, plus the Semrush, Ahrefs, Oxylabs and DataForSEO YouTube operations. Mirrors the Hive GTM Growth subscription: every operation here is subscription-only on the gateway, so a key without the plan gets 402 on all of them. 163 operations; brief + how to search and call. Optional task: what you want to find out. |
| brand_mentions_across_social | One pass over the three public social sources for a brand, product or topic: the recent posts, where the conversation is, and what is missing. Sets the source order, the page budget and when to stop; it does not score sentiment for you. |
| competitor_teardown | From one domain to a market map: how big its traffic is, which sites compete for the same audience, where that audience lives, who the search-side rivals are (often a different list — say so), and how the top competitors measure on the same yardstick. Sets the call order, the budget and the data-vintage rules; the judgement stays with the reader. |
| creator_profile_snapshot | Audience, activity and recent content of one person or brand account on X and Instagram, from the compact digests first and the raw feeds only when needed. Sets the order and the page budget; it does not judge brand fit. |
| seo_ai_visibility | Measure a brand or domain's presence in ChatGPT, Gemini and Perplexity answers: how often it is mentioned, on which prompts, which sources the models cite instead, and what the models say today when asked directly. Sets the call order and the budget; it does not write the positioning. |
| seo_keyword_opportunity | Decide whether a keyword is worth going after: volume and difficulty, the related terms that are easier, and what is actually on page one now. Sets the call order and the stopping rules; it does not write the content. |
| seo_site_audit | A defensible first-pass SEO health check of one domain: authority, organic footprint, backlink profile, top competitors and page speed. Sets the call order, when to cross-check two providers and when to stop; it does not decide what to fix first. |
| enrich_lead_list | Turn a list of emails or names into people with titles, companies and verified contact details, plus the company facts behind them — using the bulk endpoints so one call does ten. Sets the batch size and the credit rules; it does not write anything back to Apollo. |
| find_people_at_company | From a company domain to a shortlist of named people with verified work emails: enrich the company, search people by title, then match each shortlisted person for contact details. Sets the order and the credit budget; it does not write to Apollo (no contacts, no sequences). |
| crypto_market_health | Market-wide read: the largest coins and their moves, what is trending, which sectors are up, and how Bitcoin traded over the last week. Sets the order and the size of each call; it does not predict. |
| look_up_token_market | Resolve an ambiguous ticker, name or contract address into a single CoinGecko asset, then pull its current numbers and — only if asked — history and venue liquidity. Handles the two traps that make CoinGecko answers wrong rather than merely incomplete: ticker symbols are not unique, and half the endpoints want a CoinGecko id while the other half want a chain plus a contract address. |
| compare_event_odds | Look up what prediction markets currently imply about an event, checking Polymarket against Kalshi so a single venue's pricing is never the only evidence. Handles the two venues' incompatible identifiers and filters, and says plainly when only one side has a market. |
| track_topic_on_x | Assemble a defensible picture of what X/Twitter is saying about a topic: which posts, from whom, how the conversation is shaped, and what is missing. Sets the call order and the stopping conditions; it does not decide what the chatter means. |
| research_brief | A sourced answer to one question: two complementary searches, the full text of the few pages that matter, and a short brief with citations. Sets the order, the page budget and when a synthesis model is worth its price; it does not replace reading the sources. |
| search_youtube_deliberately | Run a YouTube search where the locale and the filter token are chosen on purpose rather than left to chance, and report the result set in a way the reader can audit. Covers the two parameters that quietly change what comes back. |
| company_fundamentals | Resolve a company, then pull the slices of its financial record that the question actually needs — price, statements, ratios, earnings, insiders, filings — without dragging back everything. Encodes the two things that reliably go wrong here: half these tools refuse to answer without a period argument, and several pairs look interchangeable but differ by an order of magnitude in how much they return. |
| stock_snapshot_brief | Price, valuation, latest earnings, what analysts expect, who inside is buying or selling, and the news behind it — six calls, one page. Sets the order and where to stop; it does not give investment advice. |
| stock_chatter_workflow | Decide between the one-call bundle and the underlying atomic tools when looking at what X/Twitter is saying about stocks, then read the result honestly — including which sources failed and what the mention counts do and do not mean. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| venue_field_map | Field-by-field mapping between Polymarket and Kalshi for the concepts that do not translate cleanly: identifiers, pagination, time filters, and status. Read this before querying both venues for the same question — three of the differences (boolean vs enum status, ISO vs Unix timestamps, offset vs cursor paging) produce wrong results rather than errors. |
| topic_report_template | A section-by-section skeleton for writing up X/Twitter topic research so a reader can audit it: what was searched, what came back, who said it, and what is missing. Read this when the deliverable is a written report rather than raw posts. Pairs with the track_topic_on_x prompt, which covers how to gather the material. |
| reading_the_bundle | What each field in the twitter_stock_pulse response actually means, and the three ways it is commonly misread: treating `mentions` as a ranking, missing a partial failure recorded in `coverage`, and not noticing that one tool call billed several upstream calls. Read once before interpreting the bundle; it does not change between calls. |
TDQS
Scored across 5 tools
Each tool has a clearly distinct role: list_categories covers category-level discovery, search finds operations by task, get_details exposes contracts, and use/batch_use execute single or batched operations. There is no meaningful overlap between any pair.
Names are all lowercase imperative commands with underscores, and the pair use/batch_use is logical. The main inconsistency is that search and use are bare verbs while list_categories and get_details include a noun object, but the pattern remains predictable and readable.
Five tools is well-scoped for a gateway server that fronts a huge catalogue: two discovery tools, one detail tool, and two execution tools. Every tool earns its place and the count is appropriate.
The server covers the full lifecycle of working with the AIsa catalogue: browse categories, search operations, inspect contracts, execute one operation, and execute many operations. No obvious dead ends or missing capabilities for the stated purpose.