JulianPatty · Sep 23, 2026 · Sep 23, 2026
diff --git a/CONTENT-ALIGNMENT.md b/CONTENT-ALIGNMENT.md
@@ -0,0 +1,79 @@
+# Documentation alignment audit — September 23, 2026
+
+This audit compares the local documentation with the current Innflow application source, including universal search, authoring contracts, execution handlers, navigation, and subscription limits. Application revision at review: `5ab764a564e371633b9cfc4166d84d69353a559c`. Documentation base: `43cc54675dffc41308dc89d5316062dac90312f1`.
+
+## Implemented changes
+
+Every item below has been implemented in this checkout. Paths under `src/` refer to the sibling Innflow application; `content/` paths refer to this documentation repository.
+
+| # | Mismatch or gap | Documentation correction | Application evidence |
+| --- | --- | --- | --- |
+| 1 | No complete reference tied to universal search | Added a generated reference covering all 94 standard node entries and 309 exact operation IDs | `src/components/universal-search-items.ts`, `src/config/node-registry.ts`, `src/config/tool-operations-registry.ts` |
+| 2 | Handwritten configuration summaries omitted many fields | Added complete shared configuration field lists, selectors, conditional-field notes, and secret-field guidance | `src/features/editor/lib/authoring/contracts.ts`, `src/tool-configs/` |
+| 3 | No repeatable catalog drift check | Added `docs:sync` and `docs:check`; stale reference or operation lists fail verification | Same registry and authoring-contract sources as items 1–2 |
+| 4 | Innflow Trigger had no dedicated guide | Added purpose, typed input, caller selection, and links to trigger/action configuration | `src/features/copilot/lib/trigger-authoring.ts`, `src/features/triggers/lib/entrypoint.ts` |
+| 5 | Trigger guide required manual independent-trigger enablement | Replaced with automatic draft upgrade on second-trigger save; retained busy/deployed restrictions | `src/features/triggers/server/automatic-execution-version.ts`, `src/features/editor/hooks/use-add-node.ts` |
+| 6 | Nodes/workflows/deployment lesson repeated retired trigger setup | Updated all three entry-point explanations; documented selected-path execution | `src/features/triggers/lib/entrypoint.ts`, `src/inngest/utils.ts` |
+| 7 | Manual Trigger claimed to have no output variable/data | Documented variable-picker and active-trigger metadata instead of denying output | `src/features/triggers/lib/entrypoint.ts`, `src/features/copilot/lib/trigger-authoring.ts` |
+| 8 | Condition guide described JavaScript expressions as the new-node configuration | Documented typed comparisons, AND/OR combinations, and fixed If/Else outputs; retained a legacy-mode explanation | `src/features/tools/components/condition/executor.ts`, `src/features/editor/lib/authoring/contracts.ts` |
+| 9 | Branching lesson implied LLM Condition was the default current form | Distinguished typed new Conditions from saved legacy LLM configurations | Same evidence as item 8 |
+| 10 | Router guide omitted first-match semantics | Described best-match and first-match routing, ordered rules, and Else fallback | `src/features/tools/components/router/executor.ts` |
+| 11 | Code guide promised the last expression always becomes output | Documented expression/stdout parsing, Python JSON printing, result/log wrappers, and declared output fields | `src/features/tools/components/code/executor.ts`, `src/features/tools/components/code/lib/code-utils.ts` |
+| 12 | HTTP Request guide used the wrong picker name and output paths | Added API label, all seven methods, auth/body choices, milliseconds timeout, and `httpResponse` paths | `src/tool-configs/http-request/block.ts`, `src/features/tools/components/http-request/executor.ts` |
+| 13 | HTTP failures were not distinguished from HTTP status responses | Documented structured non-2xx responses versus transport/configuration failures | `src/features/tools/components/http-request/server/api-request.ts` |
+| 14 | Transform Array described one generic expression for all operations | Documented map expressions, typed filters, aggregations, partial-result behavior, and result/count output | `src/features/tools/components/transform-array/types.ts`, `executor.ts` |
+| 15 | Sleep guide omitted the current Wait label and date mode | Added duration/date-time modes and output timestamps | `src/features/tools/components/sleep/executor.ts`, `src/config/node-registry.ts` |
+| 16 | Knowledge Base node documented search only | Added Upload, file fields, asynchronous processing, score threshold and correct search output paths | `src/features/tools/components/knowledge-base/executor.ts` |
+| 17 | Knowledge Base feature page had an extra output wrapper | Corrected `kb_results.results` and explained upload readiness | Same evidence as item 16 |
+| 18 | Variable examples taught a universal output prefix | Replaced with verified node-specific examples and qualified field-specific type handling | API, Code, Knowledge Base, Tavily, Label Fields and Transform Array executors |
+| 19 | Quickstart repeated the universal output-prefix assumption | Uses a Tables example and directs readers to actual output schemas | `src/features/tools/components/tables/`, `src/features/editor/lib/authoring/specialized-outputs.ts` |
+| 20 | Label Fields examples used `output.name` | Corrected to `extractedData.name` and `extractedData.email` | `src/features/tools/components/label-fields/executor.ts` |
+| 21 | Tavily examples used `output.results` | Corrected to direct `results` | `src/features/tools/components/tavily-search/executor.ts` |
+| 22 | AI Agent guide assumed a fixed output-text wrapper | Directs readers to the configured agent result and variable picker | `src/features/tools/components/ai-agent/executor.ts` |
+| 23 | OpenRouter guide made an unmaintained numeric model-count claim | Defers available choices to the node's current selector | `src/config/node-registry.ts`, `src/features/copilot/lib/node-type-catalog.ts` |
+| 24 | Copilot guide advertised per-message model selection | Clarified automatic Copilot model policy versus workflow AI-node model selectors | `src/features/copilot/server/model-experiment.ts`, `src/features/copilot/hooks/use-copilot-chat-go.ts` |
+| 25 | Assistant page advertised a separate active chat runtime | Rewrote as retained settings; explained Home redirects and current Copilot usage | `src/app/(dashboard)/(rest)/assistant/page.tsx`, `chat/page.tsx`, `ARCHITECTURE.md` |
+| 26 | Landing page repeated the active-Assistant claim | Updated the card to retained Assistant settings | Same evidence as item 25 |
+| 27 | Edit Fields and tool-only nodes were described as standard new-node choices | Marked legacy/hidden visibility and gave supported authoring alternatives | `src/config/node-registry.ts`, `src/components/universal-search-items.ts` |
+| 28 | Meta was presented as the picker choice | Uses Facebook/Instagram labels while retaining Meta connection terminology | `src/config/node-registry.ts` |
+| 29 | Slack Create Canvas was presented as working | Explicitly identifies it as searchable but rejected by Innflow's current implementation; avoids claiming Slack lacks a public API | `src/tools/slack/actions/create-canvas.ts`, `src/features/tools/components/slack/executor.ts` |
+| 30 | Slack configuration table listed only messaging operations | Included Create Canvas with the availability limitation above | Slack operation enum, block, and action sources |
+| 31 | Calendar guide omitted List Calendars and Word omitted Write | Expanded exact operation lists and Word description | `src/config/tool-operations-registry.ts`, Google Calendar/Microsoft Word node configuration |
+| 32 | Numerous integration guides showed partial or outdated operation names | Aligned 53 mapped guide sections with the exact search catalog; added complete reference links | `scripts/tool-operation-audit.json` records the old lists; `scripts/tool-guide-map.json` records each mapping |
+| 33 | Basic was used as a current plan label | Updated workflow retention, execution logs, and tables to Pro | `src/config/pricing-tiers.ts`, `src/features/subscriptions/lib/subscription-limits.ts` |
+| 34 | Team table claimed obsolete 5–15 seats and Free single-seat limits | Documented current workspace/invitation limits and separated them from billing seats | Same evidence as item 33 |
+| 35 | Credentials instructions depended on an outdated settings hierarchy | Linked the actual `/credentials` application route | `src/features/credentials/components/credentials.tsx` |
+| 36 | MCP guide used a generic output wrapper | Documented direct `content`/optional `structuredContent`, agentic `text`, and all four execution modes | `src/features/tools/components/mcp/executor.ts` (the older field tooltip still incorrectly says direct `result`) |
+| 37 | Google Forms summary claimed metadata retrieval | Limited the action description to its actual Get Responses operation | `src/config/tool-operations-registry.ts`, Google Forms action configuration |
+
+## Operation-by-operation change list
+
+`scripts/tool-operation-audit.json` captures each of the 53 affected guide sections, its node type, previous operation text, and resulting operation count. It distinguishes an incomplete existing list from a previously missing list. The current exact values are in `scripts/tool-catalog.snapshot.json`.
+
+Two source hints need runtime interpretation: the MCP field tooltip still advertises a direct `result` wrapper, and the Code tooltip overgeneralizes expression capture. The docs generator uses executor-backed wording for those fields rather than repeating the stale hints. This documentation task does not change those application tooltips.
+
+The generated reference intentionally lists discoverability separately from operational readiness. Slack Create Canvas is the known discoverable-but-unimplemented action found in this review. AI Decision is documented separately as a gated capability. An operation appearing in search is not evidence of a successful external provider execution.
+
+## Maintaining alignment
+
+From this documentation repository, with the application dependencies installed in the sibling `innflow` checkout:
+
+```sh
+npm run docs:sync
+npm run docs:check
+npm run build
+npm run verify
+```
+
+Set `INNFLOW_SOURCE=/absolute/path/to/innflow` if the application checkout is elsewhere. The sync reads source definitions without contacting providers, databases, or production. Review its diff; do not automatically publish changes. The check compares the generated reference with fresh application definitions and all 53 mapped operation lists. It also checks regression claims for triggers, output paths, Copilot and Assistant. The standalone verifier checks rendered pages, navigation, local links and fragments, every catalog anchor, search and 404 behavior.
+
+The snapshot uses standard picker availability (AI Decision disabled), not production environment flags. Conditional fields and live selector options still require editor inspection. This change does not provision a vector index, ingest public MDX into Copilot, run provider actions, deploy the docs, or modify application behavior.
+
+## Verification
+
+- `npm run docs:check` passed: 94 standard nodes, 309 exact operation IDs, 53 mapped guide operation lists, and the regression claims.
+- `npm run build` passed, including TypeScript checking and MDX compilation.
+- `npm run verify` passed against the built standalone server: 60 content pages, navigation, 81 local targets, local fragments, all 94 reference anchors, Innflow Trigger search, Loop search, and unknown-route 404 behavior.
+- `git diff --check` passed.
+- 28 existing content pages changed and one complete reference page was added. The audit contains 37 implemented changes plus the 53-section operation change inventory.
+- No deployment, provider execution, database change, or vector indexing was performed. These are local documentation and rendered-site results, not production verification.
diff --git a/CONTENT-REVIEW.md b/CONTENT-REVIEW.md
@@ -10,3 +10,13 @@ Reviewed against `innflow-ai/innflow` origin/main commit `7c55d9e58e9927b9f728a7
 - Free allowance discrepancy remains unresolved: latest origin/main specifies 500 credits, while the public pricing page advertises 1,200. Docs refer readers to their account allowance instead of choosing an unverified live value. Removed the old $5 and 400-credit claims.
 
 This is a focused review, not an accuracy audit of all documentation. The custom domain was transferred to the replacement on 2026-09-18; Mintlify now tracks the preserved archive/mintlify branch.
+
+
+## 2026-09-23: tool catalog and behavior alignment
+
+See [CONTENT-ALIGNMENT.md](CONTENT-ALIGNMENT.md) for the larger change list and
+source evidence. The update adds a generated universal-search reference, aligns
+53 guide operation lists, and corrects trigger setup, execution semantics, output
+paths, Copilot/Assistant behavior, hidden nodes, and plan terminology. Slack
+Create Canvas is explicitly documented as currently rejected by Innflow's
+implementation. Source coverage is distinct from live provider validation.
diff --git a/README.md b/README.md
@@ -55,3 +55,23 @@ docker compose up -d --build
 The unprivileged Node container runs the standalone Next.js server on `127.0.0.1:8080` through the existing Compose port mapping. Place your HTTPS reverse proxy in front of it. Search requires the server runtime; do not serve this build with the previous static Nginx configuration.
 
 The default `npm start` command also runs the built application from a complete checkout. For a standalone bundle, copy `public/` and `.next/static/` into their corresponding locations under `.next/standalone/`, then run its `server.js`.
+
+
+## Keep tool documentation aligned with Innflow
+
+See [the September 23 alignment audit](CONTENT-ALIGNMENT.md) for corrected behavior,
+source evidence, and the operation-level change list. With the application checkout
+at `../innflow` (or `INNFLOW_SOURCE` set to its path), run:
+
+```sh
+npm run docs:sync
+npm run docs:check
+npm run build
+npm run verify
+```
+
+`docs:sync` regenerates the complete tool reference and source snapshot.
+`docs:check` compares them with the current application and checks the operation
+lists in the handwritten guides. If a guide fails, update that guide to match the
+new operation contract before publishing. Review generated changes as well as
+handwritten guidance; a listed operation does not guarantee provider readiness.
diff --git a/content/docs/academy/branching.mdx b/content/docs/academy/branching.mdx
@@ -14,7 +14,7 @@ A Condition node branches based on configured conditions. Use rule-based conditi
 
 ## Use AI deliberately
 
-The Condition configuration also supports LLM-based evaluation. This can help when a decision depends on the meaning of text, but it needs examples and review. Do not use a model judgment in place of access controls.
+New Condition nodes use typed rules and fixed If/Else branches. Use Router when a decision depends on text meaning. Earlier workflows may still contain LLM-based Condition settings; that legacy form is not the default for new nodes.
 
 AI Decision is a separate capability with availability controls. If it is enabled for your workspace, read its [reference](/tools/ai-decision) before building around its Decided and Needs review outputs.
 

diff --git a/content/docs/academy/deployment.mdx b/content/docs/academy/deployment.mdx
@@ -18,7 +18,7 @@ Check required fields, selected credentials, branch destinations, and sample out
 
 ## Trigger availability
 
-Workflows start from a configured trigger. Before adding another trigger to a legacy workflow, use **Triggers → Enable independent triggers** and follow the conversion review. The capability must also be enabled in the environment. If the menu is missing or the action is unavailable, contact support or your administrator. See [adding another trigger](/triggers#adding-another-trigger) for the conversion steps and the exact error message.
+Add additional triggers directly. Existing drafts switch to independent execution when the second trigger saves; no separate enablement setting is needed. Undeploy and drain active or pending runs before changing entry points. See [adding another trigger](/triggers#adding-another-trigger), then test every intended entry point before redeploying.
 
 ## Try it
 

diff --git a/content/docs/essentials/integrations.mdx b/content/docs/essentials/integrations.mdx
@@ -9,7 +9,7 @@ Innflow integrates with dozens of external services. Depending on the service, a
 
 <Steps>
   <Step title="Go to Credentials">
-    Navigate to **Settings → Credentials** from the sidebar.
+    Navigate to [Credentials](https://app.innflow.ai/credentials) from the sidebar.
   </Step>
   <Step title="Connect a service">
     Click **New Credential** and choose the service. OAuth integrations redirect you to authorize Innflow; API-based integrations ask for the required credential. Platform-managed integrations do not need a workspace credential.

diff --git a/content/docs/essentials/nodes.mdx b/content/docs/essentials/nodes.mdx
@@ -7,13 +7,14 @@ Nodes are the building blocks of every workflow. Each node performs a specific a
 
 ## Trigger Nodes
 
-A trigger starts a workflow run. Legacy workflows cannot add a second trigger until independent triggers are enabled, and multiple-trigger authoring must also be available in the environment. See [adding another trigger](/triggers#adding-another-trigger) for setup details.
+A trigger starts a workflow run. Add more triggers directly; saving a second trigger upgrades an existing draft automatically. Each event starts its selected path. See [adding another trigger](/triggers#adding-another-trigger).
 
 <Columns cols={3}>
   <Card title="Manual" icon="play" href="/triggers#manual-trigger">Run on demand</Card>
   <Card title="Webhook" icon="webhook" href="/triggers#webhook-trigger">HTTP requests</Card>
   <Card title="Chat" icon="message-square" href="/triggers#chat-trigger">Conversational</Card>
   <Card title="Schedule" icon="calendar-clock" href="/triggers#schedule-trigger">Recurring</Card>
+  <Card title="Innflow" icon="workflow" href="/triggers#innflow-trigger">Called by another workflow</Card>
   <Card title="Input" icon="text-cursor-input" href="/triggers#input-trigger">Custom form</Card>
   <Card title="Form" icon="file-text" href="/triggers#form-trigger">Public hosted form</Card>
   <Card title="Stripe" icon="credit-card" href="/triggers#stripe-trigger">Payment events</Card>
@@ -69,6 +70,8 @@ Action nodes do the work in your workflow. They are organized by category:
   </Card>
 </Columns>
 
+The [complete tool and operation reference](/tools/reference) lists every standard universal-search entry and its configuration fields. Search **API** for HTTP requests and **Wait** for duration or date-based pauses.
+
 ## Node Configuration
 
 Click or double click any node to open its configuration panel. Common fields include:
@@ -89,7 +92,7 @@ Nodes that change external systems can create real side effects during testing.
 | **Wait** | Pause for a duration or until a specific time |
 | **Human Approval** | Pause until a workspace member approves or rejects |
 | **Wait for Event** | Pause until an external callback arrives |
-| **Edit Fields** | Reshape an object without writing code |
+| **Edit Fields** | Legacy node retained in existing workflows; use Code or Transform Array for new workflows |
 | **Transform Array** | Map, filter, or reduce a list |
 
 ## Node Status