docs: add Inventor credit, conversation, and feedback docs

[NiamhRelevance]

What

Adds content the canonical Inventor page (build/invent/invent.mdx) was missing, reconciled from the superseded consolidate/inventor branch and rebased onto current main. All additions match the page's existing component patterns (Tabs, CardGroups, Steps).

• Credit usage — new section after "How to use Inventor" explaining the real-time session credit counter, with a screenshot (images/invent-credit-counter.png).
• Conversations — new tab in "Inventor features" covering Chats session history (sessions persist across close/refresh and can be resumed from the Chats dropdown).
• Feedback — new tab in "Inventor features" covering thumbs up/down ratings and the manual thumbs-down bug-report popup.

Notes

• Scoped to build/invent/invent.mdx plus the one new image — no other pages touched.
• Deliberately omits automatic bug logging, which main already documents via the "Log bugs to engineering" card and warning (avoids duplication).

🤖 Generated with Claude Code

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
relevanceai	🟢 Ready	View Preview	Jun 18, 2026, 3:05 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[github-actions]

🎯 Vibe check

Reviewed: 1 file (1 with issues, 0 clean) + 1 image (no review needed)

Scores

	Dimension	Score	What's holding it back
🟢	Consistency	8/10	"agent" and "tool" in new content at lines 145 and 207 should be capitalized as product features per CLAUDE.md. Note: the entire existing file has this pattern, so it's pre-existing — but the new content repeats it. No banned words, no British spellings.
🟡	Technical clarity	7/10	"current session" (line 147) is ambiguous. The Conversations tab (line 193) says opening the panel starts a new chat — readers will reasonably wonder whether the credit counter resets per chat. It doesn't say.
🟢	Non-technical clarity	9/10	Clear, accessible language. No jargon dumps. The Feedback tab explains the thumbs-up/down controls without assuming prior knowledge.
🟢	Structure	9/10	New sections slot in naturally. Minor CardGroup misuse in the Feedback tab — two one-sentence definitions dressed up as feature cards.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: Clean, practical additions that fill a real gap — credit tracking and session persistence were missing and readers needed them. The one thing worth fixing before ship is the "current session" ambiguity in the credit counter section: pair it with the session lifecycle info from the Conversations tab so readers know when the counter resets.

🔧 Issues (3)

• build/invent/invent.mdx:145 — "generating an agent" → "generating an Agent". Refers to the Relevance AI product feature, should capitalize. Same pattern at "requesting an adjustment" is fine (adjustment isn't a product term). Note: the entire file uses lowercase for agent/tool/workforce — this pre-existing inconsistency with CLAUDE.md is worth a separate cleanup pass, but the new content shouldn't add to it.
• build/invent/invent.mdx:207 — "the generated agent or tool didn't match what you described" → "the generated Agent or Tool didn't match what you described". These are Relevance AI product features being named.
• build/invent/invent.mdx:147 — "tracks only the credits used in your current session" — "current session" is undefined here. The Conversations tab (line 193) says "Opening the panel starts a new chat" — so sessions are per-chat. Does the counter reset each time you open the panel? Readers building up cost across many sessions need to know. Add a one-liner: e.g. "The counter resets when you start a new chat session."

🧩 Component suggestions (1)

• build/invent/invent.mdx:202–209 — The <CardGroup cols={2}> for thumbs-up/thumbs-down in the Feedback tab contains two cards that are each one sentence long and have no href. They're not navigable choices — they're a two-item glossary. CLAUDE.md says CardGroup isn't appropriate for short descriptor lists. Replace with prose: "Thumbs up marks the response as accurate and helpful; thumbs down marks it as incorrect, unhelpful, or a generated Agent or Tool that didn't match your description." This keeps the meaning and drops the visual overhead.

⚠️ Contradictions (1)

• Internal ambiguity (not a cross-file contradiction, but worth flagging): build/invent/invent.mdx:147 says the credit counter tracks credits "in your current session," while build/invent/invent.mdx:193 says each panel open starts a new chat. These two pieces of information together imply the counter resets per chat, but the credit section doesn't say that explicitly. A reader checking their costs across multiple chats may not realize they need to sum across sessions. Clarify in the credit section.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	2
Total lines processed	~324

Files read: build/invent/invent.mdx (235 lines), build/introduction.mdx (50 lines), build/agents/create-an-agent.mdx (91 lines)

[github-actions]

🎯 Vibe check

Reviewed: 1 file (1 with issues, 0 clean)

Note on non-doc file: images/invent-credit-counter.png is a new image referenced at build/invent/invent.mdx:146. It supports the credit usage section and has no independent documentation impact.

Scores

	Dimension	Score	What's holding it back
🟡	Consistency	7/10	Product feature names (Agent, Tool, Workforce, Trigger) appear lowercase in multiple places across the page; "agent builder" and "agent editor" should be capitalized per the proper-noun rule
🟡	Technical clarity	8/10	Model default not stated; Settings access step missing what to actually click
🟢	Non-technical clarity	9/10	Well-explained throughout; MCP jargon introduced on line 15 but fully resolved in the MCP vs Inventor section
🟡	Structure	8/10	No closing CTA on a concept page

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: Solid feature page with good component choices, no banned words, and clear examples throughout. The main friction is inconsistent capitalization of product feature names — "agents, tools, workforces, and triggers" appear lowercase through much of the page when they should be capitalized as Relevance AI proper nouns. Two vague UI references could also leave users stuck.

🔧 Issues (7)

• build/invent/invent.mdx:4 — front matter description uses "agents, tools, workforces, and triggers" (all lowercase) → "Agents, Tools, Workforces, and Triggers"; product feature proper nouns per CLAUDE.md must always be capitalized
• build/invent/invent.mdx:21–43 — Card body text uses lowercase "agents, tools, workforces, and triggers" throughout; capitalize when referring to Relevance AI product features (e.g. "Create new Agents, Tools, Workforces, and Triggers")
• build/invent/invent.mdx:58 — "tool, agent, workforce, knowledge, and home" → "Tool, Agent, Workforce, Knowledge, and home" (the list refers to specific Relevance AI builder pages; "home" is a generic nav label, not a product feature, so lowercase is fine)
• build/invent/invent.mdx:74 — "the agent builder" → "the Agent builder" (refers to the product-specific editing interface)
• build/invent/invent.mdx:83,86 — "the agent editor" → "the Agent editor" (same reasoning; the create-an-agent.mdx image alt text uses "Agent builder" capitalized, confirming this is the intended form)
• build/invent/invent.mdx:163 — "there's a default if you don't change it" — name the default model, or at minimum write "the current default is shown in the dropdown" so users know what to expect
• build/invent/invent.mdx:190–192 — Step 2 ("Click Settings") says "The settings pop up" without specifying what the user actually clicks to open them. Name the specific button or icon (e.g. "Click the gear icon in the panel header" or "Click Settings").

🧩 Component suggestions (1)

• build/invent/invent.mdx:205–211 — <CardGroup cols={2}> with only 2 items (Thumbs up / Thumbs down) is below the threshold where visual weight is warranted. These are simple two-option descriptions, not navigable choices. A two-item bullet list or a single sentence ("Use thumbs up when the response was accurate, thumbs down when it was incorrect or unhelpful") reads more naturally and saves the space.

🏗️ Page structure (1)

• build/invent/invent.mdx — no closing CTA. This is a concept page for a significant platform feature and should end with a clear direction. The final <Tip> is informational but not actionable. Add a ## What's next? section pointing to 2–3 natural follow-on pages, for example: /build/agents/create-an-agent, /build/tools/create-a-tool, and /integrations/mcp/programmatic-gtm/mcp-server.

⚠️ Contradictions (2)

• build/invent/invent.mdx:99–100 says to start Inventor from the Agents page by clicking Build with Inventor in the header. build/agents/create-an-agent.mdx:41–43 says to click + Create Agent and then select Invent. If these describe the same flow, the button label needs to be reconciled across both pages. If they're genuinely two separate entry points, both docs should acknowledge that the other path exists.
• build/invent/invent.mdx:113–115 says to start Inventor from the Tools page by clicking Build with Inventor in the header. build/tools/create-a-tool.mdx:22–24 says to click + New Tool and then select Invent. Same issue — different button names for what appears to be the same flow.

🔋 Credit usage

Item	Count
Files reviewed	1 MDX + 1 image
Context pages read	4
Total lines processed	~517

Files read: build/invent/invent.mdx (244 lines), build/agents/create-an-agent.mdx (91 lines), build/tools/create-a-tool.mdx (102 lines), integrations/mcp/programmatic-gtm/mcp-server.mdx (40 lines, partial), integrations/mcp/mcp-client.mdx (40 lines, partial)

[NiamhRelevance]

Why the "Documentation Lint Checks" failure is expected

The failing check is the Supademo embed lint rule, not a broken link — the link-rot and Mintlify deployment checks both pass. It's flagging the "How to use Inventor" demo in build/invent/invent.mdx:

Supademo embed isn't using the standard wrapper [paddingTop: '56.25%' missing from wrapper <div>]
Supademo embed is missing the purple border / rounded corners

This is intentional and we're leaving it as-is. The rule expects the standard 16:9 wrapper (paddingTop: '56.25%'). This embed deliberately uses padding-top: 44% instead, because this particular Supademo recording is not 16:9 — its native aspect ratio is wider/shorter. Forcing the standard wrapper letterboxes and crops the demo, which is exactly what the earlier docs: restore Supademo aspect ratio commit set out to fix. Conforming to the snippet would pass the check but visibly break the demo.

(For the record, the purple border and rounded corners are present on the iframe — the rule flags them too because it validates against the standard wrapper as a single block, and our wrapper deviates.)

Two things worth noting:

• This embed is pre-existing on main — this PR didn't add or touch it. The lint only surfaces it because the check scans the whole changed file, and this PR edits invent.mdx for unrelated content (credit usage, version restore, feedback).
• The separate afdocs check failure is also pre-existing and unrelated — it's about site-wide llms.txt structure / markdown parity on other pages (e.g. /get-started/pricing), nothing in this PR.

[github-actions]

🎯 Vibe check

Reviewed: 1 file (1 with issues, 0 clean). images/invent-credit-counter.png is a new screenshot added for the credit usage section — no content issues.

Scores

	Dimension	Score	What's holding it back
🟡	Consistency	6/10	Product-term capitalization is wrong in 8 distinct locations: "agents, tools, workforces, and triggers" appears throughout in lowercase instead of "Agents, Tools, Workforces, and Triggers". Additionally, the product is called "Inventor" but lines 129 and 163 call it "Invent".
🟡	Technical clarity	8/10	UI references are specific and steps are well-formed. One factual discrepancy: the MCP Client page lists preset connections as "Notion, Canva, or Atlassian" but this page says "Notion and Linear".
🟡	Non-technical clarity	8/10	"MCP" appears at line 15 without explanation before it's defined in the later section. "LLM" in the Model selection tab (line 163) is used without definition. Both are minor since MCP is explained later and "LLM" is in an optional tab, but worth softening for non-technical readers.
🟢	Structure	9/10	Logical flow from concept → capabilities → access → usage → features → comparison. Missing a closing CTA for an overview page — readers land here, learn what Inventor is, then have no directed next step.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

✨ Overall vibe: The page is well-structured and the prose is clear — good use of Tabs, Steps, and CardGroup throughout. The main problem is systemic: Relevance AI product terms (Agent, Tool, Workforce, Trigger, Knowledge) are lowercased in at least eight places across the page, and the product's own name appears as "Invent" twice when it's "Inventor". Fix the capitalization pass and close the contradiction about MCP client presets, and this is a solid overview page.

🔧 Issues (11)

• build/invent/invent.mdx:8 — "invent an agent" and "invent a tool" → "invent an Agent" and "invent a Tool" (product-feature capitalization)
• build/invent/invent.mdx:21 — "agents, tools, workforces, and triggers" → "Agents, Tools, Workforces, and Triggers"
• build/invent/invent.mdx:39 — "logs and tools" → "logs and Tools" (refers to Relevance AI Tools)
• build/invent/invent.mdx:58 — "a persistent side panel from any builder page — tool, agent, workforce, knowledge, and home" → "Tool, Agent, Workforce, Knowledge, and Home"
• build/invent/invent.mdx:129 — "Invent is expensive to run" → "Inventor is expensive to run" (the product is called Inventor, not Invent)
• build/invent/invent.mdx:137 — "agents, tools, workforces, and triggers" → "Agents, Tools, Workforces, and Triggers"
• build/invent/invent.mdx:163 — "your Invent session" → "your Inventor session"
• build/invent/invent.mdx:227 — "builds or edits agents, tools, and workforces for you" → "Agents, Tools, and Workforces"
• build/invent/invent.mdx:232 — "Call your existing Relevance AI tools from external clients" → "Relevance AI Tools"; "let your agents call services" → "let your Agents call services"
• build/invent/invent.mdx:239 — "build your agents and tools with Inventor" → "Agents and Tools"
• build/invent/invent.mdx:163 — "there's a default if you don't change it" — which model is the default? Technical readers will want to know. Name it or link to where it's configurable.

🧩 Component suggestions (1)

• build/invent/invent.mdx:205–211 — The Feedback tab uses a two-card CardGroup for "Thumbs up" / "Thumbs down". These aren't navigable and carry only one-line descriptions each — the exact pattern CLAUDE.md flags as inappropriate for CardGroup. Replace with two bullets: - **Thumbs up** — the response was accurate and helpful. / - **Thumbs down** — the response was incorrect, unhelpful, or the generated Agent or Tool didn't match what you described.

🏗️ Page structure (1)

• build/invent/invent.mdx — No closing CTA. This is a concept/overview page; readers who finish it need a directed next step. Add a ## What's next? section with links to /build/agents/create-an-agent (create your first Agent using Inventor) and /integrations/mcp/programmatic-gtm/mcp-server (call your tools from external clients after building with Inventor).

⚠️ Contradictions (1)

• build/invent/invent.mdx:233 says Inventor's MCP Client lets "your agents call services like Notion and Linear" — but integrations/mcp/mcp-client.mdx:25 lists preset connections as "Notion, Canva, or Atlassian". Linear is not mentioned anywhere on the MCP Client page, not even as a custom connection example. Either update the example services to match the presets ("Notion and Canva"), or add a parenthetical clarifying Linear requires a custom connection.

🔋 Credit usage

Item	Count
Files reviewed	1
Context pages read	4
Total lines processed	~534

Files read: build/invent/invent.mdx (241 lines), build/agents/create-an-agent.mdx (91 lines), build/tools/create-a-tool.mdx (102 lines), integrations/mcp/programmatic-gtm/mcp-server.mdx (60 lines), integrations/mcp/mcp-client.mdx (40 lines)