Jinba

What it is

Jinba is an AI workflow automation platform for enterprise teams. It has two parts: Jinba Flow, "for building and automating workflows", and Jinba App, "for executing them". Technical and semi-technical builders design workflows in Flow, and business users run them in App through chat or auto-generated forms. The company behind the site is named as Carnot AI, Inc., the site is offered in English and Japanese, and it carries a "Backed by Combinator" badge. The pages do not give a company location beyond a playful "Built in San Fransokyo" line in the footer. Its homepage headline is "Enterprise AI Workflows. Without the Enterprise Wait."

What you build and how

You start in Jinba Flow by describing the workflow in natural language. The homepage walks through three steps. First you describe: "Say what you want to build." Then you test, running the workflow "instantly with real data" and debugging in real time. Then you share, deploying it as an API or an MCP server so your whole team can use it. The enterprise page adds batch processes to that list, so a workflow can be an API, a batch process or an MCP server. The site's sample prompt is "Send weekly metrics to Slack every Monday at 9am", and the demo test log shows steps such as input validation, fetching from a Sheets API, processing 24 rows and sending to a #sales channel.

After the chat step, the generated workflow can be refined in a visual editor with a flowchart interface and detailed configuration. The FAQ calls Flow "a vibe-coding platform that combines ease of no-code use", and says you can build in natural language, edit detailed settings visually and deploy instantly to production. Deployed workflows are available as MCPs from Jinba App, "with no need for complex UI development or maintenance."

Jinba App is the execution side. The FAQ puts it plainly: "Builders create in Flow, users consume in App." Business users interact through a chat interface or auto-generated forms, which the enterprise page frames as a guardrailed UI that keeps processes inside the defined rules. This separation of building from running is the product's main design idea, and it suits organizations that want many people to use a workflow but few to change it.

The sample use case on the enterprise page is a sales opportunity approval process. When an opportunity exceeds $100,000 it routes automatically for VP approval, deals over $500,000 also need review from the finance team, each step keeps audit logs, and data is validated against Salesforce records before routing. The homepage's featured use case is loan screening for financial institutions: design the screening in natural language, set criteria from business rules, configure conditional routing by loan type, amount, risk level and regulatory requirements, validate applications and enrich them with data from credit bureaus and internal databases, and keep audit trails with timestamps and reviewer details.

For AI models the site shows private hosting through AWS Bedrock, Azure OpenAI or your own models, with example entries for Claude 3.5 Sonnet, GPT-4 Turbo, Meta Llama 3 and a self-hosted 70B model. Hosting can be on-premises or in a private cloud. The site labels one option "On-Prem Jinba Private" and says you can "Run Jinba in your own environment."

Integrations are listed as "100+ pre-built integrations plus custom connectors for your internal systems." The logo strip names OpenAI, Gmail, Slack, HubSpot, Salesforce, Notion, Linear, GitHub, Teams and Dropbox. Custom connectors come with "Dedicated Engineer Support": the page shows a sample chat in which a customer asks for a connector to an internal CRM, shares the API spec, and gets the connector the next day. Organization management covers spaces, roles and approvals for your team.

Evidence from the pages

Security is the center of the claims. The homepage says "SOC 2 compliant, with end-to-end encryption, SSO, RBAC, audit logging, and on-premise deployment options in place." The enterprise page says "SOC II compliance" and describes the platform as built for Fortune 500 organizations. Its FAQ states that audit trails capture initiators, approvers, timestamps and contextual information, and the homepage adds full change history. The data FAQ says "Your data never leaves your region." Treat these as the company's own statements.

Pricing is not published as numbers on these pages. The FAQ says pricing "scales with your team size and usage", each plan includes a specific number of team members, workspaces and workflow executions, and there is an enterprise plan for teammate management features. Credits "are used for workflow execution", and you can move to a higher tier for more credits. The homepage has "Try it for free" and "Contact Sales" buttons, and the header offers a free assessment and a demo.

Onboarding and support are described in the FAQ. You can sign up for a free account and begin creating flows through chat. The company says it provides templates and guided tutorials, plus training sessions and dedicated customer success managers for enterprise customers. Support is tiered: community support for free users, priority chat support for paid plans, and dedicated success managers with priority support for enterprise customers.

One customer quote on the homepage credits "Quick integration with existing MCPs and the ability to deploy agents on the Jinba App" for streamlining biomedical data analysis. The enterprise page lists a "Backed by Combinator" badge and a Fortune 500 positioning but does not name customers in the text.

Who it suits and who it does not

Jinba suits an IT or operations group at a large or regulated organization that wants business teams to use AI workflows without waiting on an IT backlog, and that needs audit logging, role-based access and the option to host on its own infrastructure. The loan screening and approval-routing examples show the target: multi-step processes with thresholds, reviewers and records. Teams that cannot send data to a public model provider will also look at the private hosting options through Bedrock, Azure or a self-hosted model.

It is a heavier choice for a solo founder or a small team that needs one quick automation. The product is organized around spaces, roles, approvals and dedicated support, and the pages state no prices, so a small buyer cannot judge cost from the text. The split between Flow and App also means someone has to act as the builder before anyone else can run a workflow. The homepage says a free account is available to try, which is the sensible way to test whether chat-based generation produces what you meant before involving a sales conversation.