--- title: "Totalum Integration" description: "Totalum is an AI app builder that turns a natural-language brief into a deployed Next.js application with a managed database, auth, payments and hosting. Agents launch and update projects, drive the build agent, query and edit database records, deploy to production, manage domains and files and react to project events." url: https://flowrunner.ai/integrations/totalum date_modified: 2026-09-04T14:59:52-07:00 --- # Totalum [Developer Tools](https://flowrunner.ai/integrations/category/developer-infrastructure) Totalum is an AI app builder that turns a natural-language brief into a deployed Next.js application with a managed database, auth, payments and hosting. Agents launch and update projects, drive the build agent, query and edit database records, deploy to production, manage domains and files and react to project events. [Verified](https://flowrunner.ai/integrations/verified "What does verified mean?") · 1 trigger · 59 actions · API key · available [Totalum website](https://totalum.app/) · [Platform Documentation](https://www.totalum.app/docs/api/overview) · Capability data verified 2026-08-27 1. A change request for a customer-facing internal tool is approved 2. Run AI Agent sends the prompt to the project's build agent, which edits real source and rebuilds 3. On Totalum Event fires when the agent finishes, so the flow waits instead of polling for half an hour 4. Get Project re-reads which preview URL is currently valid, because that answer changes between runs 5. Get Agent Conversation and Get Logs return what the agent did and what the build reported 6. The preview link, the change summary, and any warnings the run returned are posted for review 7. The owner reads the warnings and the diff before Deploy To Production publishes the change ## What This Integration Enables Totalum turns a written brief into a deployed Next.js application with a managed database, auth, payments, and hosting attached. What makes it worth connecting rather than just using is that the whole lifecycle is addressable. An agent can launch a project, prompt it into shape, query and edit the rows inside it, read its file tree, ship it to production, and put a customer's own domain in front of it. That is a genuinely different proposition from a code generator, because the thing that comes out the other end is running software with an address. FlowRunner's contribution here is not speed. Totalum is already fast. It is that the build loop has consequences a fast loop tends to hide. Everything is credit-metered, so a runaway prompt cycle spends real money. Only one heavy operation runs per project at a time, so a deploy fired while the agent is still working is refused rather than queued. A successful launch can still carry a warnings array describing steps that did not complete. And several operations replace rather than merge. Agents handle the polling, the status reading, and the warning collection reliably; the calls that overwrite somebody's work are the ones this connector deliberately routes through a person, which is what [human-in-the-loop](https://flowrunner.ai/concepts/human-in-the-loop) means in a build pipeline rather than an approval queue. ### Without FlowRunner **Small tools never get built**: The internal request queue fills with two-day applications that never reach the top of an engineering backlog **Prototypes stall at the demo**: An app exists in a builder, and shipping it to a real URL with real data is a separate project **Nobody can see what the AI changed**: A generated application is a black box, so reviewing it means reading the whole thing again ### With FlowRunner **Building is a step in a flow**: A brief becomes a running application through the same flow that handles the rest of the process **The path to production is one call**: Deploy, custom domain, logs, and rollback are all operations a workflow can drive **Every prompt leaves a record**: The agent conversation, the version snapshots, and the file-level diffs are all readable before anything ships ## Use Case Scenarios ### A request in the backlog becomes a running application An internal tool request is triaged in [Linear](https://flowrunner.ai/integrations/linear) and marked approved. The agent calls Launch Project, which creates the project and starts the build in one call, then waits on the On Totalum Event trigger instead of polling Get Agent Status for half an hour. When the event arrives, the agent reads Get Project for the currently valid preview URL, pulls Get Agent Conversation for what was actually asked and built, and checks Get Logs for anything the build complained about. All of it posts to [Slack](https://flowrunner.ai/integrations/slack) as a single review card with the preview link. If the requester wants changes, Run AI Agent takes the follow-up prompt against the same project, and the loop repeats without anyone opening a builder. ### The application's own data becomes part of the wider process A shipped Totalum app is now collecting records. The agent calls Get Database Structure to learn the tables and their real names, then Query Database on a schedule to pull the new rows. Records that need enrichment are updated with Edit Database Record, related records are joined with Link Records, and a rolled-up view is appended to [Google Sheets](https://flowrunner.ai/integrations/google-sheets) for the team that does not have a login. Where the app collects a lead, the record is mirrored into [HubSpot](https://flowrunner.ai/integrations/hubspot) so the sales process sees it. The generated application stops being an island and becomes another system in the flow. ### A designer's file and an engineer's repository both reach the same project The project is connected to a repository through Connect GitHub and to a design account through Connect Figma, so the build agent can work from a linked design rather than a description. When an engineer edits the repository directly, the agent does not sync on its own. It reads Get GitHub Status for the connected branch, List Versions and Get Version Diff for what the build agent changed inside Totalum since the last sync, and presents both sides. Only after a person names the authoritative side does Pull From GitHub run. Secrets follow the same discipline: Create Secret adds them per environment, and Download Environment Variables is treated as handling a credential bundle rather than reading a config file. ## Human-in-Loop Highlight Totalum's GitHub surface has five operations and only one of them writes: Pull From GitHub, which overwrites the Totalum project with the repository's code. There is no push. So an agent handed the instruction to sync a project with its repository has exactly one option available, and it moves in the direction that destroys the build agent's work rather than the direction that publishes it. Version snapshots make that survivable rather than safe, and recovering one is itself a heavy operation that discards everything since. So the agent gathers and stops. It posts: "This project and its repository have both changed since the last sync. Totalum has 3 agent versions since then, touching 11 files including the checkout page. The repository branch is 2 commits ahead, touching 4 files. Pull From GitHub will overwrite the Totalum side. Which one is authoritative?" The owner answers, and the agent runs the pull, polls Get Pull Status, and triggers the rebuild. Reading both sides of a divergence is work a machine should do. Deciding which version of the truth survives is not. Agent processes routinely Detects exception requiring judgment Clear match Continues automatically Ambiguous Routes to human via preferred channel Human decides Agent resumes with decision ## Agent Capabilities 59 actions ### Account and Credits 4 - **Get Account** Returns the credit balance, split into plan credits and purchased ones. Every operation is credit metered and a call needs a positive balance, so this is the check before a flow that runs the agent repeatedly. - **Get Credit Costs** Returns what each operation costs. Running the agent and deploying are the expensive ones, which is what makes a build loop worth budgeting before it starts. - **Get Spending Analytics** Returns consumption over time and by operation, so it is visible which projects and which operations are doing the spending. - **Update Credit Limits** Sets a project's monthly credit caps by category. The control that stops one runaway project consuming the whole account's allowance. ### Projects 6 - **Launch Project** Creates a project and starts the AI agent building it in one call. The recommended entry point for anything new, and asynchronous by nature. A success here means the project exists, not that every step ran, so the warnings it returns are read rather than discarded. - **Create Project** Creates an empty project without starting the agent. Used only when the agent will be run separately. - **List Projects** Returns projects with paging, search, sorting, group filtering, and a creation date range. - **Get Project** Returns a project with its status, server state, preview URLs, production URL, and custom domain. It also names which preview URL is currently valid, and that answer changes, so it is re-read after every agent run rather than cached. - **Update Project** Updates a project's label, description, or group. The project id itself is permanent. - **Delete Project** Permanently deletes a project along with its database, deployments, and source. Only allowed for API plan projects, so a project created in the web interface may not be removable through this connector at all. ### The Build Agent 5 - **Run AI Agent** Sends a prompt to the project's build agent, which edits the real source and rebuilds. The same operation drives the first build and every change after it. - **Get Agent Status** Returns whether the agent is still working, and carries the running conversation so progress can be shown while it works. - **Get Agent Conversation** Returns the complete history between the flow and the agent, across every prompt. The audit record of what was asked and what came back. - **Stop Agent** Signals a running agent to stop. Credits already spent are not refunded and the project is left wherever the agent had reached, so this is a containment action rather than an undo. - **Start Or Restart Server** Starts or restarts the project's development server. Totalum archives an idle server and charges to auto-start it when needed, so warming it deliberately is cheaper than paying for it mid-deploy. ### Database 7 - **Get Database Structure** Returns a project's tables with their fields, types, and relations. Called before any other database operation, because the value the write operations want is the table's real name rather than its display label. - **Query Database** Reads records with filtering, projection, sorting, and custom aggregation. - **Create Database Record** Inserts a record into a table. The id is generated and returned rather than supplied. - **Edit Database Record** Updates a record, writing only the properties supplied and leaving the rest untouched. - **Delete Database Record** Permanently deletes a record. - **Link Records** Creates a relation between two records through a reference property. Relations are set this way rather than by writing a foreign key directly. - **Unlink Records** Removes a relation. Both records survive and only the link goes. ### Deployment 5 - **Deploy To Production** Builds the project and publishes it to its production URL. Asynchronous, and the operation that makes a change real for the people using the app. - **Get Deployment Status** Returns the state of the current deployment, and on success the public address to read back. - **Add Custom Domain** Attaches a customer hostname to the production deployment and returns the DNS records to configure. The domain is not live until those records resolve. - **Remove Custom Domain** Detaches the custom domain. The site stays reachable at its Totalum URL, and anyone using the custom hostname stops being served. - **Get Logs** Returns server logs from either the development or the production server. The first place to look when a deployment succeeds and the app misbehaves anyway. ### Source Code 7 - **Get Project Tree** Returns the project's file tree. Totalum generates real Next.js and TypeScript source, so this is a genuine directory listing rather than a model of one. - **Get File Content** Returns one file's contents by path. - **Write File Content** Overwrites one file entirely with the supplied content. There is no append or patch, and an empty string empties the file, so a flow reads before it writes. - **Upload File** Uploads a file into the project from a URL, so an image or asset can be added for the agent to work from. - **Download Source Code** Downloads the project's complete source as an archive. Totalum's position is that the generated code belongs to the account, and this is the operation that makes that concrete. - **Rebuild Project** Rebuilds the project after a direct file edit. Writing a file does not rebuild the app on its own, so this is the step that makes a manual change take effect. - **Get Rebuild Status** Returns the state of a rebuild, including the build errors when it fails, which is where a broken manual edit surfaces. ### Versions 3 - **List Versions** Returns the project's saved snapshots. Totalum snapshots as the agent works, so these are the points a bad prompt can be rolled back to. - **Get Version Diff** Returns the file-level differences a version introduced. Read before recovering, so what a rollback would remove is known rather than assumed. - **Recover Version** Rolls the project back to a saved version, discarding the work done since. A heavy operation that cannot run alongside a build or a deployment. ### Project Transfer 2 - **Export Project** Exports a project so it can be imported elsewhere. How a project is duplicated or turned into a starting point for the next one. - **Import Project** Imports a previously exported project into an existing one. The target is written into, so it should be a project created for the purpose. ### Project Groups 5 - **Create Project Group** Creates a group to file projects under, which is how a large account stays navigable. - **List Project Groups** Returns the groups on the account. - **Get Project Group** Returns one group with its name and the projects filed under it. - **Update Project Group** Renames a group or changes its settings. - **Delete Project Group** Deletes a group. The projects in it survive and become ungrouped. ### Secrets and Environment 3 - **Create Secret** Stores an environment variable for a project, encrypted, and scoped to a chosen environment. - **Delete Secret** Removes a stored environment variable. Anything reading it starts seeing it unset at the next rebuild or deployment. - **Download Environment Variables** Returns the project's environment variables as a file, in plaintext. The result is a credential bundle and is handled as one. ### GitHub 5 - **Connect GitHub** Connects a project to a repository so its generated source can be synchronized. The route out of Totalum's own hosting. - **Disconnect GitHub** Disconnects a project from its repository. The repository and its contents are untouched and only the link stops. - **Get GitHub Status** Returns whether a project is connected and which repository and branch it syncs with. - **Pull From GitHub** Pulls the connected repository's code into the Totalum project, overwriting what is there. The only write in this group, and the direction that loses agent work. The operation this page's human gate exists for. - **Get Pull Status** Returns the state of a pull, so the rebuild that follows it is not started early. ### Figma 4 - **Connect Figma** Connects a Figma account to a project so the build agent can read linked designs and build from them rather than from a description. - **Disconnect Figma** Disconnects the Figma account. The agent stops being able to read linked designs. - **Get Figma Status** Returns whether a project has a Figma account connected. - **Validate Figma Token** Checks a token before connecting it, so a bad credential fails at setup rather than the first time the agent needs it. ### Webhooks 3 - **Register Webhook** Subscribes a URL to one Totalum event. Only one webhook may exist per event on an account, so registering is a coordinated act rather than a per-flow habit. - **List Webhooks** Returns the webhooks registered on the account with their events and destinations. The way to find out what already owns an event before trying to claim it. - **Delete Webhook** Removes a subscription, freeing its event for another. ## Triggers 1 triggers ### Event Triggers 1 - **On Totalum Event** Fires when the build agent finishes processing a prompt, or when a project reaches its credit limit. The agent event is the one that removes polling from a build flow: rather than calling for status every fifteen seconds for up to half an hour, the flow waits here and resumes when the work is done. Each delivery carries the project id, the status, and the prompt that finished. The credit limit event is the other half of the same story, because it is what tells a flow that a project has stopped being able to spend before anyone notices the builds have quietly stopped. ## Frequently Asked Questions ### What can FlowRunner do with Totalum? FlowRunner agents can run Get Account, Get Credit Costs, and Get Spending Analytics in Totalum, plus 56 more actions. ### Does connecting Totalum to FlowRunner require OAuth? No. Totalum connects to FlowRunner with an API key, no OAuth flow required. ### Can Totalum trigger a FlowRunner workflow automatically? Yes. Totalum supports 1 trigger that can start a FlowRunner workflow automatically. **Work at Totalum?** This integration exposes Totalum to AI agents on every FlowRunner plan, including through MCP, at no cost to you. [See what FlowRunner offers integration partners](https://flowrunner.ai/integrations/partners), including how to keep this page current. --- Markdown version of https://flowrunner.ai/integrations/totalum. Site index: https://flowrunner.ai/llms.txt