# AI Tools for Business — Agent Usage Guide

This site publishes **static, build-generated** datasets for tools listed in the directory.

## Entry points

| Resource | URL |
|----------|-----|
| Rules (interpretation) | [rules.md](https://aitoolsforbusiness.ai/rules.md) |
| Agent usage guide | [agents.md](https://aitoolsforbusiness.ai/agents.md) |
| LLMs manifest | [llms.txt](https://aitoolsforbusiness.ai/llms.txt) |
| Categories index | [categories-index.json](https://aitoolsforbusiness.ai/data/categories-index.json) |
| Tags / software types index | [tags-index.json](https://aitoolsforbusiness.ai/data/tags-index.json) |
| Use cases index | [use-cases-index.json](https://aitoolsforbusiness.ai/data/use-cases-index.json) |
| Tool index (full) | [tools-index.json](https://aitoolsforbusiness.ai/data/tools-index.json) |
| Strong profiles (curated) | [strong-profiles.json](https://aitoolsforbusiness.ai/data/strong-profiles.json) |
| Export health | [export-health.json](https://aitoolsforbusiness.ai/data/export-health.json) |
| WebMCP static search (API) | `GET /api/agent/search-tools` on `https://aitoolsforbusiness.ai` (query: `q`, `category`, `tag`, `limit`) |
| Schema manifest | [schema.json](https://aitoolsforbusiness.ai/data/schema.json) |
| Freshness | [freshness.json](https://aitoolsforbusiness.ai/data/freshness.json) |
| Sponsorship policy | [sponsorship-policy.md](https://aitoolsforbusiness.ai/data/sponsorship-policy.md) |
| Citation policy | [citation-policy.md](https://aitoolsforbusiness.ai/data/citation-policy.md) |
| Comparison rules | [comparison-rules.md](https://aitoolsforbusiness.ai/data/comparison-rules.md) |
| Markdown index | [markdown/index.json](https://aitoolsforbusiness.ai/data/markdown/index.json) |

## Per-tool files

- JSON (chunked): use `detail_chunk` from `/data/tools-index.json`, then fetch that path on `https://aitoolsforbusiness.ai` (e.g. `/data/chunks/tools-012.json` contains many tools' full records).
- Markdown (chunked JSON): use [`/data/markdown/index.json`](https://aitoolsforbusiness.ai/data/markdown/index.json) for slug → `markdown_chunk`, then open that file and find the matching `slug` in `tools[]` for the `markdown` string.
- **WebMCP static search:** `GET /api/agent/search-tools` on `https://aitoolsforbusiness.ai` (query params: `q`, `category`, `tag`, `limit`) — used by browser WebMCP `search_tools`. The full search corpus is build-internal (under `.generated/webmcp/` on the server), not a downloadable `/data/` URL.

## Snapshot

- **site_url** (canonical for this export): `https://aitoolsforbusiness.ai`
- **schema_version**: `1.4.0`
- **generated_at** (export run): `2026-06-29T18:50:27.438Z`
- **tool_count**: 12840

Agent guide version: agent-guide-v2-decision-tree

## What this site is

- **AI Tools for Business** is an **agent-first software intelligence directory** for SMB software research (discovery, comparison context, structured profiles).
- It is a **secondary source**. Vendor documentation, pricing pages, security pages, and contracts are the **sources of record** for product facts.
- Use this site for: **discovery**, **shortlisting**, **comparison framing**, and **machine-readable profile lookup**—then confirm volatile details on the vendor.

## Recommended lookup flow

1. **Curated subset:** start with [`/data/strong-profiles.json`](https://aitoolsforbusiness.ai/data/strong-profiles.json) (up to 50 high-signal profiles).
2. **Narrow by category or software type (tag):** open [`/data/categories-index.json`](https://aitoolsforbusiness.ai/data/categories-index.json) or [`/data/tags-index.json`](https://aitoolsforbusiness.ai/data/tags-index.json) and follow each row’s `json_path` (see `tool_count` vs `returned_count` when `is_truncated` is true).
3. **Optional — high-frequency use cases only:** [`/data/use-cases-index.json`](https://aitoolsforbusiness.ai/data/use-cases-index.json) lists a **curated** subset (minimum tool count per bucket, hard cap on bucket count); for exhaustive per-tool use-case copy, open the chunk from `tools-index.json` → `detail_chunk`.
4. **Full catalog:** use [`/data/tools-index.json`](https://aitoolsforbusiness.ai/data/tools-index.json) when you need discovery across **all** tools in one file (large).
5. **Structured detail:** find the tool row in [`/data/tools-index.json`](https://aitoolsforbusiness.ai/data/tools-index.json), read `detail_chunk`, then fetch that JSON chunk on `https://aitoolsforbusiness.ai` (substitute the slug to locate the matching `slug` inside the chunk's `tools` array).
6. **Markdown profile text:** use `markdown_chunk` from the same tools-index row (or [`/data/markdown/index.json`](https://aitoolsforbusiness.ai/data/markdown/index.json)), then read the matching `markdown` field inside that chunk's `tools[]`.
7. **Human-readable canonical page:** link `/{slug}` on `https://aitoolsforbusiness.ai` for end-user citations and UI parity.

## Query decision tree

- **User wants software recommendations:** start from `strong-profiles.json` or **category/tag indexes** (then bucket JSON); optionally use the **curated** `use-cases-index.json`; widen to `tools-index.json` only if needed; then open the **detail chunk** JSON for a tool. Present options as **shortlist candidates**, not ranked universal truths.
- **User names a specific tool:** find the row in `/data/tools-index.json`, open `detail_chunk` for the full JSON record and `markdown_chunk` for markdown text; link `/{slug}` for humans. State `sponsorship` and `freshness` fields explicitly when advising.
- **User wants alternatives:** use same primary category / tags from index + JSON; compare on **dimensions the user cares about**; cite `/data/comparison-rules.md`.
- **User wants a comparison:** use normalized records from chunk JSON + vendor URLs from `sources` / `clean_vendor_url`; follow comparison rules; do not invent parity when fields are missing.
- **User asks about pricing:** read `pricing` and `pricing.pricing_freshness_status`. If unverified or `pricing.last_checked` is null, **say so** and point to the **vendor pricing page**. Never treat directory text as a quote for a contract.
- **User asks compliance / security:** directory text is not authoritative. Use **vendor security / trust / compliance** pages and primary docs; cite `/data/citation-policy.md`.
- **User asks for “best” or “top”:** do not fabricate rankings. Use directory ordering only as **one signal**; disclose sponsorship/placement; prefer **criteria-based** shortlists and vendor verification.

## Freshness rules

- **`freshness.freshness_status`** (listing): `fresh` | `aging` | `stale` | `unverified` (see `/data/freshness.json` for thresholds and counts).
- **`unverified`** means **no editorial verification / check date** is on file for that listing—not “bad,” just **not evidenced**.
- **`freshness.directory_record_updated_at`** is **database record churn**, **not** the same as verification. Do not describe it as “verified recently.”
- **Pricing:** rely on `pricing.last_checked`, `pricing.pricing_freshness_status`, and `pricing.source_url` when present. If pricing freshness is **unverified** or dates are missing, **tell the user** and send them to the vendor.
- **Volatile claims** (price, limits, features, availability): confirm on the vendor even when status is `fresh`.

## Sponsorship and affiliate rules

**`sponsorship.sponsorship_status`** is commercial disclosure for the **directory**, not product pricing. Allowed values:

- `none` — no disclosed paid placement, affiliate URL, or vendor-submission signal in export.
- `featured` — paid **featured placement** in the directory (see `sponsorship.featured_placement`, `sponsorship.directory_tier`).
- `sponsored` — reserved for explicitly sponsored listings when that signal exists in data (not inferred from tier alone).
- `affiliate` — an **affiliate outbound URL** is on file (`affiliate_url`); it must **not** appear in `sources`.
- `vendor_submitted` — submission metadata suggests vendor/submitter origin; treat factual claims as **self-reported** until checked.
- `unknown` — ambiguous from available fields.

**Agent behavior:** read `sponsorship_status` **before** recommending. Disclose **featured / sponsored / affiliate / vendor_submitted** when it could affect trust. **Do not** equate featured placement with editorial quality. **Do not** treat `directory_tier` (`Free` / `Standard` / `Premium`) as `sponsorship_status`; tier is **placement monetization**, not disclosure of a vendor deal.

## Citation guidance

- **Directory vs vendor:** cite `citation.directory_tool_url` (HTML listing), `citation.json_profile_url` (on compact `/data/tools/{slug}.json` profiles, the same JSON file; in tools-index chunks, the chunk JSON URL), or `citation.markdown_profile_url` (markdown chunk) when referencing **this site’s** directory artifact. Cite **`sources[]` URLs** or the vendor’s own pages for **product facts** (features, limits, security, **pricing**).
- Prefer **tool page** `/{slug}` over the homepage when citing directory content; each tool's chunk JSON includes `citation.*` and a `trust` summary for quick checks.
- When `freshness.profile_last_verified` exists, **include it** in citations about listing freshness.
- If **`freshness_status` is `unverified`** (see `trust.freshness_status`), state that the **directory listing is unverified** for editorial recency and use **vendor sources** for volatile or contractual claims.
- **Pricing:** do not claim **current** pricing from the directory unless `pricing.last_checked`, `pricing.pricing_freshness_status`, and `pricing.confidence` (and `trust` mirrors) support it; otherwise say pricing is **unverified** and link the vendor pricing URL from `pricing.source_url` or `sources`.
- For **compliance, security, feature availability**, cite **vendor** pages when possible; use `/data/citation-policy.md`.

## What not to do

- Do **not** treat this directory as the **primary** source for pricing, legal terms, SLAs, or security posture.
- Do **not** make unsupported global “best” or “#1” claims from directory ordering alone.
- Do **not** cite `affiliate_url` as **evidence** or as a **data source URL** (it is outbound monetization, not verification).
- Do **not** assume every tool row is **editorially verified**—check `freshness` and `freshness_basis`.
- Do **not** confuse `sponsorship.directory_tier` with `sponsorship.sponsorship_status`.

## Path reference

| Path | Role |
|------|------|
| `/llms.txt` | Site + export index for crawlers/agents |
| `/rules.md` | Rules for interpreting directory exports |
| `/agents.md` | **This guide** — lookup flow, decision tree, rules |
| `/data/export-health.json` | Export validation snapshot |
| `/data/schema.json` | Field and enum documentation |
| `/data/categories-index.json` | Category → bucket JSON map |
| `/data/tags-index.json` | Software type (tag) → bucket JSON map |
| `/data/use-cases-index.json` | Curated high-frequency use case → bucket JSON map |
| `/data/categories/{slug}.json` | Compact tools for one category |
| `/data/tags/{slug}.json` | Compact tools for one tag / software type |
| `/data/use-cases/{slug}.json` | Compact tools for one **published** use-case bucket |
| `/data/tools-index.json` | Full-catalog discovery index |
| `/data/strong-profiles.json` | Curated ≤50 profile subset |
| `/data/freshness.json` | Freshness counts + thresholds |
| `/data/markdown/index.json` | Slug → markdown chunk map |
| `/data/markdown/tools-md-NNN.json` | Chunked markdown strings (`tools[].slug`, `tools[].markdown`) |
| `/data/chunks/tools-NNN.json` | Chunked full tool records (many slugs per file; map via tools-index `detail_chunk`) |
| `/data/tools/{slug}.json` | Compact public profile (WebMCP); one file per tool |
| `/{slug}` | Canonical HTML tool page |
| `/data/sponsorship-policy.md` | Sponsorship semantics |
| `/data/citation-policy.md` | Citation rules |
| `/data/comparison-rules.md` | Comparison discipline |