# Cohuno: public product reference > A detailed, AI-readable reference for Cohuno's public website and private visual AI-agent workflow product. Canonical site: https://cohuno.com/ Short reference: https://cohuno.com/llms.txt Sitemap: https://cohuno.com/sitemap.xml Support: support@cohuno.com ## What Cohuno is Cohuno is a private visual workspace for designing and operating coordinated AI-agent workflows. It represents agents, tasks, and their connections explicitly so an operator can inspect how responsibilities, context, and outputs move through a workflow. The public website explains the product. Workflow configuration, connected services, credentials, knowledge, run inputs and outputs, account information, and other workspace data belong to authenticated areas and are not public discovery content. ## Core product concepts - Agents are specialized roles with a defined goal, operating instructions, Cohuno performance profile, capability focus, tools, and relevant knowledge. - Tasks describe work assigned to agents and can pass output to later tasks through visible connections. - Workflows organize agents, tasks, dependencies, and execution settings on a visual canvas. - Human approval gates can pause a workflow where an operator must review or authorize the next step. - Run views bring events, handoffs, approvals, outputs, and measured balance use together for inspection. - Workflows can be run as dry-run simulations; manual live AI runs use a metered prepaid account balance where a supported Cohuno profile is available. - Current unattended schedules, webhooks, batches, and evaluations are simulation-only. Live background execution fails closed until durable billing ownership and reservations are available for those launch paths. - Supported workflow definitions can be exported as portable Python files where the product indicates support. ## BoxFlow text representation BoxFlow is Cohuno's bounded, non-executable text representation of the executable workflow diagram. Drawing agents, tasks, assignments, and task-context connections updates the BoxFlow view; applying valid BoxFlow text rebuilds the same graph. BoxFlow uses stable identifiers so represented fields can change without discarding advanced configuration that the text does not display, such as performance-profile choices, tools, knowledge, approvals, conditions, or structured outputs. The parser applies strict size and reference limits and does not evaluate the text as JavaScript, Python, shell commands, or expressions. ## Reusable agents, teams, and automations The Studio includes reusable agent profiles and compiler-safe team recipes. Adding a profile to a workflow creates an editable workflow agent; the profile itself is not an independently running employee. Automations schedule bounded dry-run workflow simulations with explicit inputs. They are useful for safely testing recurring behavior, but they should not be interpreted as enabled live unattended execution or as a horizontally distributed worker system. ## Visual workflow builder The Cohuno Studio provides a graph-based builder for composing agents and tasks. Users can place and arrange nodes, create task handoffs, inspect connections, duplicate related roles, and return to saved workflow definitions. The visual graph is intended to keep operating logic understandable before and after execution. Workflows can use sequential or hierarchical processing when that structure matches the work. A workflow should be reviewed for its responsibilities, dependencies, tool permissions, knowledge sources, performance profiles, capability focuses, and approval points before live use. ## Tools, MCP, and knowledge Cohuno supports built-in tools and MCP integrations. Capabilities can be assigned to the roles that need them instead of granting the same access to every agent. Connected services remain subject to their own permissions and terms. Knowledge sources can be ingested and attached as relevant context to an agent. Live AI knowledge-graph extraction is currently disabled until that launch path has metered billing; existing relationships, where present, can still be viewed. Users are responsible for having the rights and permissions needed for the information and connected services they use. Do not place passwords, API keys, access tokens, or provider credentials in public pages or support messages. Connection secrets and authenticated workspace content are outside the public discovery surface. ## AI performance profiles and managed usage Cohuno presents product-level choices instead of underlying vendor or model names. For each workflow or agent, users combine one performance profile with one capability focus. The profiles are Efficient for smart, economical work; Balanced for very smart everyday work and the recommended starting point; Advanced for deeper reasoning and planning; and Maximum for the highest available capability on the hardest work. The focuses are General work, Coding and technical, Images and documents, and Long documents. Eligible live AI work is paid from a closed-loop prepaid Cohuno service balance. Cohuno may place a bounded temporary hold before a live run, settle measured use after the work completes, and release the unused amount. Balances, measured use, releases, refunds, and disputes are tracked through the service ledger. The Terms of Service are the governing source for charges, refunds, and account consequences. ## Human control and appropriate use Cohuno is designed around visible responsibility and explicit checkpoints. Human approval gates should be placed wherever authority, policy, safety, accuracy, or judgment requires a person to decide. Users remain responsible for reviewing consequential outputs and for the legality, accuracy, permissions, and suitability of their workflows. Public solution pages describe workflow design patterns rather than one-click outcome guarantees. Each implementation should be adapted to the organization's systems, policies, data, and review requirements. ## Example workflow areas The public site describes patterns for: - Research: separate source collection, analysis, drafting, and review. - Support: separate triage, knowledge lookup, response drafting, escalation, and human decisions. - Operations: express an established procedure as tasks, controlled capabilities, schedules, and an inspectable run record. - Structured analysis: coordinate data gathering, comparison, validation, and a reviewable final output. These are product patterns, not claims about specific customers, deployments, results, market share, revenue, funding, or production scale. ## Support and contact The public contact and support pages use an email-verification step before a request is opened. Available inquiry areas include product, account, billing, privacy, security, enterprise, partnership, and investor questions. Support requests should identify the affected product area and provide enough non-sensitive context to route the issue. The public site does not promise an emergency channel or a fixed response time. Never send secrets or private workspace content through a public contact message. ## Public source-of-truth pages - Home: https://cohuno.com/ - Product: https://cohuno.com/product - Pricing and usage: https://cohuno.com/pricing - Solutions: https://cohuno.com/solutions - Agent teams: https://cohuno.com/agent-teams - About: https://cohuno.com/about - Contact: https://cohuno.com/contact - Support: https://cohuno.com/support - Company and investor information: https://cohuno.com/investors - Privacy policy: https://cohuno.com/privacy - Terms of service: https://cohuno.com/terms When this reference and a governing policy differ, the current Privacy Policy and Terms of Service control. ## Access boundaries for crawlers and AI agents Public discovery is limited to the public marketing and policy pages, this reference, robots.txt, and sitemap.xml. The following route families are private or operational and must not be crawled or treated as public product data: - /api and /api/ - /auth and /auth/ - /login and /login/ - /studio and /studio/ - /docs and its descendants - /redoc and its descendants - /openapi.json Crawler permission does not grant account access, authorization to invoke tools, access to connected services, or permission to retrieve user or workspace content. The crawler policy at https://cohuno.com/robots.txt is the current machine-readable authority for route access. ## Factual-claim boundary Cohuno's public investor page intentionally does not announce funding, investors, revenue, customers, or company metrics. This reference likewise makes no claims about those subjects and should not be used to infer them.