Free Intro Class|BotSpot Builder Program|Learn. Build. Earn.|Master AI Agents • Build Real-World Projects • Industry-Ready Portfolio Limited Seats Available
BotSpot · developer docs
Workflow node reference
Every node the workflow builder offers, one page each — what it is for, how it behaves at run time, every configuration field, and what usually goes wrong. Ports and configuration are generated from the runtime registry, so they match what the builder actually serves.
- nodes
- 49 nodes
- categories
- 16 categories
- builders
- 4 builders
- with side effects
- 9 with side effects
Start here
A workflow is a graph of typed steps wired together on a canvas. Most edges carry data; a flow edge carries nothing and only sequences; a tool edge binds a capability to a model so it can decide to call it. Wiring a tool with a data edge is the most common first mistake.
Two nodes route — Condition and Switch / Case fire exactly one branch and prune the rest. Everything else computes. Logic, transforms and filters are free and deterministic; cost enters when a node reaches a model, calls a purchased tool, or invokes another workflow.
Input
2 nodes- Big Text Inputinput.bigtextA named, long-form field the run is started with — a résumé, a job description, a transcript — rather than a chat turn.cost none
- Chat Inputinput.chatWhere a conversation enters the graph. It takes whatever the channel handed the run and turns it into the three shapes the rest of the canvas expects.cost none
AI
4 nodes- Agent Loopagent.loopGive the model a goal and let it work. It plans, calls the tools bound to it, reads the results and keeps going — saving its progress after every turn, so a long run survives a restart.cost model tokens
- LLMai.llmOne model call. Give it messages, get an answer — and if you bind tools to it with a tool edge, the same node becomes an agentic loop that calls them until it can answer.cost model tokens
- LLM with Memoryai.llm_memoryA conversational LLM that remembers. It fetches the session's prior turns, answers with them in context, and writes the new turn back — three nodes fused into one.cost model tokens
- Prompt Templateai.prompt_templateLoad an approved prompt from PromptCenter by id and version, bind its variables, and hand the finished messages to a model.cost none
Logic
2 nodes- ConditionifRoute on a single comparison. It reads a value, compares it, and forwards the data out either the true or the false branch.cost none
- Switch / Caseswitch-caseRoute on one value across many branches. Each case label you add becomes its own output port; anything unmatched leaves through default.cost none
Transform
7 nodes- Combinetransform.combineMerge several structured outputs into one. Point several nodes at it and it fans them in.cost none
- Compose Messagetransform.compose_messageAssemble the canonical messages array — system, then history, then the current user turn — from three independent sources.cost none
- Extract Listtransform.extractPull one field out of every item in a list, into a flat list of values. A pure extractor — it does not match, compare or branch.cost none
- Find Rowtransform.lookupFind the first row in a list whose field matches a value, and keep the whole row. This is the membership question and the answer to who they are, in one node.cost none
- Radar Datatransform.radar_dataReshape several scored LLM outputs into the series a radar chart needs.cost none
- Save Variablevariable.setStore a value under a name for the rest of the run, and read it anywhere by that name — no data edge needed to each consumer.cost none
- Valuetransform.valueEmit a fixed value — a literal, or something read from the run. Its output does not depend on its input; wiring an input just triggers it.cost none
Memory
6 nodes- Append to Documentdocument.appendAdd a line to one of the business's documents. Safe when several people are talking to the agent at once — every append lands.cost none
- Config Fetchconfig.fetchLoad this org's configuration — business name, contact line, industry — and hand it to a Prompt Template as variables.cost none
- Documentation Fetchdocument.fetchGround the agent in the business's own Document Space, so it answers from documented facts rather than improvising.cost none
- Knowledge Fetchknowledge.fetchSemantic search over the org's pre-ingested knowledge, injected as grounding context before the model answers.cost model tokens
- Memory Fetchmemory.fetchPull the session's prior turns out of the memory service and inject them into the messages array.cost none
- Memory Writememory.writePersist this turn — the user's message and the assistant's reply — back to the memory service. It acts rather than computes, so it is triggered by a flow edge.cost none
Knowledge
5 nodes- Knowledge Chunkknowledge.chunkSplit crawled pages into embedding-sized pieces, each carrying where it came from and a hash of its text.cost none
- Knowledge Crawlknowledge.crawlFetch every same-host page reachable from a starting URL, breadth-first, up to a hard page cap. The front of the ingestion chain.cost none
- Knowledge Embedknowledge.embedAttach an embedding vector to every chunk. This is the step that costs model money in the ingestion chain.cost model tokens
- Knowledge Filterknowledge.filterDrop boilerplate before you pay to process it — blog archives, tag pages, pagination, login screens.cost none
- Knowledge Storeknowledge.storeWrite the embedded chunks into a knowledge namespace, replacing whatever those same pages stored last time.cost none
Data
3 nodes- Data Read (SQL)data.readRun a read-only SELECT against a connected database and get rows back, with run values bound as named parameters.cost none
- Spreadsheet Rowssheet.readRead an uploaded spreadsheet as rows — one record per line, keyed by column heading, ready for Filter and Fan out.cost none
- Update Spreadsheet Rowsheet.writeWrite one value into an uploaded spreadsheet: find the row whose key column matches, and set another column on it.cost none
Integrations
2 nodes- Notionnotion.readRead from Notion at a fixed point in the graph — fetch a page, search the workspace, or query a database. A typed data node, not a tool.cost whatever the tool charges
- Notion Writenotion.writeUpdate a Notion page or database row's properties. The deterministic write counterpart to the Notion read node.cost whatever the tool charges
Tools
2 nodes- Function Toolfunction.toolUse a tool you purchased from the marketplace. Connect it to an LLM or Agent Loop with a tool edge so the model can call it on its own.cost whatever the tool charges
- Tool Calltool.callInvoke a registered tool through the gateway at a fixed point in the graph. Deterministic — the graph decides when it runs, not the model.cost whatever the tool charges
Automation
3 nodes- Fan outautomation.fanoutTurn a list of rows into one agent run each. It is what makes for-each work without a loop on the canvas.cost none
- Filterautomation.filterThe node that decides who the rest of the workflow runs for. It takes a list in, tests every row against a condition you build on the node itself, and passes through only the rows that match.cost none
- Workspace Dataworkspace.readList rows from a bespoke workbench table. The batch pipeline's source when the customer's data lives in their own database.cost whatever the tool charges
Image
2 nodes- Edit Imageimage.editImage-to-image editing: an existing image plus a prompt, out comes an edited one. It chains straight after Generate Image.cost model tokens
- Generate Imageimage.generateTurn a text prompt into one or more images. Wire a reference image in and the same node edits instead of inventing.cost model tokens
Video
1 nodeAudio
1 nodeOutput
7 nodes- Avatar Videooutput.avatarAnimate an avatar image so it speaks — a lip-synced talking-head clip from text or from audio you supply.cost model tokens
- Chat Outputoutput.chatThe original chat reply node. Deprecated — Outcome does the same thing plus every other channel.cost none
- Image Outputoutput.imageShow the generated image or images as the workflow's result. A terminal node — nothing runs after it.cost none
- Outcomeoutput.outcomeThe run's outcome on a channel. Mirror replies where the message came from; naming a channel sends somewhere else.cost none
- Radar Chartoutput.radarRender a radar chart in the chat interface. A terminal node, fed by Radar Data.cost none
- Video Outputoutput.videoShow the generated clip or clips as the workflow's result. The video counterpart to Image Output.cost none
- Voice Outputoutput.voiceSpeak text aloud. It takes an upstream answer and synthesises a clip the chat surface plays.cost model tokens