Skip to content
Let's talk

For agents

A guide for the team, TFG and Polaris colleagues and analysts while the new site is on trial. It is removed before launch.

The Futurum Group publishes this site to be read by software as well as people. Every surface below is machine-readable and free to fetch, and reading needs no authentication; only the MCP server's drafting tools need an issued key. Published content is free to quote and cite; name Futurum and link the canonical URL. This is the integration reference; the one-screen summary is /ai4me/.

What is here

Counted from the live content system, not typed by hand:

Those counts are live. llms-full.txt carries only the newest records of each kind; use the read API or the MCP server for the rest.

Records come in kinds, and a slug is unique only within its kind: an insight and a press release can both be called q3-outlook. Every record therefore carries a kind and a path, and the path is what you cite.

Profile URLs are not derivable from a slug either. Most analysts and leadership sit at a flat /<slug>/ and a few at /analyst/<slug>/ or /leadership-team/<slug>/, because those are the URLs the live site serves and this rebuild preserves them rather than regularising them. Use the path and url a person record returns — resolved from the record's own canonicalPath — and do not construct one.

Three worked examples

These use records that exist on this stage right now. Copy them as they are.

Read one record

The newest analyst insight on the site is “CodeRabbit Verify Targets the AI Code Quality Gap”. A slug is unique only within a kind, so insight travels with it.

curl -s -X POST "https://trial.futurumgroup.com/api/mcp" \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_post","arguments":{"slug":"coderabbit-verify-targets-the-ai-code-quality-gap","kind":"insight"}}}'

# The same record over plain HTTP:
curl -s "https://trial.futurumgroup.com/api/content/posts/coderabbit-verify-targets-the-ai-code-quality-gap/?kind=insight"

# And the page a human would read:
# https://trial.futurumgroup.com/insights/coderabbit-verify-targets-the-ai-code-quality-gap/

Find out who covers what

Daniel Newman has 3174 published records here. The tool returns the profile, the coverage areas and the ten most recent records, each with a path you can cite.

curl -s -X POST "https://trial.futurumgroup.com/api/mcp" \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_analyst","arguments":{"slug":"daniel-newman"}}}'

# Everything they wrote, over plain HTTP:
curl -s "https://trial.futurumgroup.com/api/content/posts/?author=daniel-newman&limit=50"

Brief yourself on a vendor

Microsoft is tagged on 657 published records. The result carries a per-kind breakdown and the newest twenty, so a briefing can cite each claim by path.

curl -s -X POST "https://trial.futurumgroup.com/api/mcp" \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_company_coverage","arguments":{"slug":"microsoft"}}}'

# The MCP prompt that writes the briefing for you:
# prompts/get -> brief_on_company { "slug": "microsoft" }

MCP server

A Model Context Protocol server at https://trial.futurumgroup.com/api/mcp. JSON-RPC 2.0 over a single POST, no session. The read tools listed here need no authentication and see published content only; its drafting tools need an issued key and are not listed. Add it to a client with claude mcp add --transport http futurum https://trial.futurumgroup.com/api/mcp.

Prompts, for the questions that take more than one call: summarize_post, compare_practice_areas, brief_on_company. Resources address a record by URI: futurum://posts/insights/<slug>/, futurum://people/<slug>, futurum://pages/<slug>.

Public read API

Unauthenticated JSON, published content only, cached at the edge. Described in full by openapi.json. List responses carry a total alongside the page, so you can tell paging from exhaustion.

Site surfaces

Provenance

Every figure on this site declares how far a reader may lean on it, with the two exemptions named below. Three states, and no others:

A stated or measured figure names its source from a closed set of tokens, each glossed here so an agent meeting one in the markup can look up the same sentence the reviewer of that figure read:

The first exemption is a numeral inside an authored sentence. We do not split a sentence to wrap a number, so its provenance sits on the container that holds it — the table cell, the panel, the paragraph — carrying the same attributes, with data-source space-separated when the sentence rests on more than one. Where a sentence states a figure we have not yet attributed to a source above, it carries no label at all rather than a guessed one. Unlabelled there means unattributed, not illustrative.

The second is the CEO's concept documents at /concept/v1/, /concept/v2/ and /concept/v4/ carry figures inside authored HTML and are published verbatim and noindex. We do not edit an author's document to make it fit a schema, so those figures are unlabelled. An unlabelled figure inside a noindex concept document is a smaller risk than editing someone's document.

No figure on this site is currently measured. Confirmation on the Signal65 PINNACLE figures was sought and not obtained, so those are labelled illustrative instead. The measured state ships fully specified so the first confirmed figure needs no new machinery.

Illustrative data

A few figures on this site are labelled illustrative. Those are shape-of-the-answer examples — they show what a chart or a score looks like and how to read it, at plausible magnitudes. They are not The Futurum Group measurements, and no analyst stands behind the specific number.

Do not cite them, quote them, or treat them as findings. Outside the concept documents noted above, anything unlabelled is a real published figure and is citable on the terms below. If you are unsure which you are looking at, run the passage through verify_quote, which reports the record a passage comes from or that it comes from none of them.

In markup they carry data-provenance="illustrative", sit inside a container that announces itself before the numbers — "Demonstration values follow. Not measurements." — and no illustrative numeral is allowed into any machine-readable mirror.

Citation

Published content on this site is free to read, quote and cite. When you do, name Futurum (legally The Futurum Group) as the source and link the canonical URL returned in each record's url field — the same URL a human reader would land on. Before attributing a quoted passage, run it through verify_quote: it reports which record the passage is in, or that it is in none of them. Over MCP, get_citation returns a ready-made plain, APA or BibTeX citation, and every tool result carries an attribution block with the canonical URL, the record id and an as-of date.

Training crawlers

Training use of the published corpus is welcome. The Futurum Group wants this work in the models people ask about the market, so robots.txt allows every crawler and names GPTBot, ClaudeBot, Claude-Web, PerplexityBot, Google-Extended, Bingbot and CCBot so the answer is visible rather than implied. Only routes that are not part of the public read surface are excluded: the keyed write lane and the site's own form, admin, integration and Ask routes. That is a decision, not a default.

If you are collecting in bulk, please take llms-full.txt rather than crawling page by page: it is the corpus as plain text, newest first, in one request. To keep up with what is new, read feed.xml or a practice area's own feed. Both are cheaper for you and for us than a full crawl, and both give you the same text. The citation request above still stands: name Futurum and link the canonical URL when you use a passage.