SectorFlow
AIManage SectorFlow multi-model workspaces and chat against them from FlowRunner. Agents compare model output across providers without a connection per vendor.
What This Integration Enables
SectorFlow is not an OpenAI-compatible gateway and reading it as one will lead you into building the wrong flow. There is no model parameter on a chat request. Models are attached to a workspace, and you chat against the workspace, so a workspace holding three models answers one prompt three times. That single design decision is the product. Everywhere else in this category, multi-model means you may choose among many. Here it means several answer at once.
That makes SectorFlow the right connector for the questions where being wrong quietly is the actual risk: policy interpretation, contract language, risk classification, anything where a single fluent answer is indistinguishable from a correct one. An agent can treat convergence between models as a green light and divergence as a reason to stop, which is a far better confidence signal than any single model's self-reported certainty.
The workspace is also a governed team object rather than a private session. Create Workspace and Update Workspace control which models are attached and how chat history, context, and sharing behave, including whether a conversation is visible to just you or to the whole team. That means the record of what was asked and what each model said is a shared artifact, which is what makes this usable in work that later has to be explained to somebody. FlowRunner's contribution is the layer above: your agents run against these workspaces, and the human-in-the-loop step decides what to do when the models do not agree.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Contract clause interpretation with a dissent check
A renewal window is closing and a clause is ambiguous. The agent sends the clause and its definitions to a workspace bound to three models and receives three readings. Where all three agree, the agent records the interpretation and moves on. Where one dissents, it packages all three answers with the divergence highlighted and routes them to legal ops in Slack. Nobody is asked to review the easy clauses, and nobody misses the contested one.
Standing up a comparison workspace on demand
An evaluation starts with List Models to see which models the account can reach, then Create Workspace to bind the three candidates with team sharing enabled and history retained. The agent runs a fixed set of prompts through Send Chat Message and writes every answer to a scored table in Notion. At the end, the evaluation is a shared workspace and a table of real outputs rather than a memo describing an impression.
Changing the roster of a live workspace
A model in a production workspace is being retired. The agent calls List Workspaces to find every workspace referencing it, then proposes the replacement roster. Update Workspace swaps the models, but only after a person confirms, because every teammate and every flow chatting against that workspace gets different answers the moment the swap lands, and the prior conversation history sitting in the same workspace was produced by the old roster.
Human-in-Loop Highlight
The whole reason to run a multi-model workspace is to see disagreement, so the one thing an agent must never do here is resolve it on its own. When Send Chat Message returns three answers to a compliance question and two say the clause auto-renews while one says it does not, the agent has no defensible basis for picking a winner. Voting is not judgment, and picking the longest or most confident answer is worse. So it stops. It posts all three answers verbatim to the responsible owner, labeled by the model that produced them, with the point of divergence called out: "Three models, two readings. Models A and B read this clause as auto-renewing on 30 days notice. Model C reads the notice period as running from invoice date, which would put the deadline eleven days earlier. Which reading do we act on?" The person decides, the agent records the decision alongside the three answers in the workspace history, and the renewal is handled on a reading somebody owns. A convergent answer would have gone through without anyone being interrupted. The dissent is what earned the interruption, and detecting it is precisely what this connector is for.
Agent Capabilities
5 actionsModels
1- List Models Retrieves every large language model available to the account, with the model ID needed to build a workspace. Run this before Create Workspace to decide which models a workspace should query.
Workspaces
3- List Workspaces Retrieves all workspaces available to the account, including each workspace ID, its name, and the models attached to it. Used to find every workspace affected before a model is retired or swapped.
- Create Workspace Creates a workspace bound to one or more language models. The workspace is the unit you chat against, so attaching several models means one prompt is answered by each of them. History, context, and sharing settings decide whether the conversation is private or visible to the whole team.
- Update Workspace Updates an existing workspace: rename it, swap the models it queries, or change its history, context, and sharing settings. Supply only the fields you want to change. Swapping models changes the answers every downstream consumer of that workspace receives, so this belongs behind an approval.
Chat
1- Send Chat Message Sends a message to a workspace and returns the completion. The workspace decides which models answer, so a workspace holding several models returns a response from each one to the same prompt. Supply a Prompt for single-turn questions or a full Messages array to carry prior turns.
Frequently Asked Questions
What can FlowRunner do with SectorFlow?
FlowRunner agents can run List Models, List Workspaces, and Create Workspace in SectorFlow, plus 2 more actions.
Does connecting SectorFlow to FlowRunner require OAuth?
No. SectorFlow connects to FlowRunner with an API key, no OAuth flow required.
Can SectorFlow trigger a FlowRunner workflow automatically?
SectorFlow doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with SectorFlow
$100 in credits. No card required. Connect in minutes.