Give every customer an agent with memory, tools and a place to work.
On request or on schedule. Inside your product. Powered by ACTA.
Build an assistant that remembers preferences, works with files and returns to unfinished tasks. ACTA keeps its identity and home connected to compute.
Their preferences, projects and working history stay associated with one identity, across messages, schedules and product events.
memory/skills/files/Memory, files and configuration remain available across process restarts and sleep/wake cycles.
Wake, suspend and restart an agent through the API. Keep future work in a persisted wake schedule.
Your company defines the skills and boundaries. Each customer gets an agent with its own continuing context.
“Compare three project-management tools for our small team.”
Recall the team's size and preference for simple workflows.
Compare product pages from three approved sources.
Assemble a comparison brief in the customer's workspace.
Wait for the customer to ask for another brief.
Local example only. No settings are saved or deployed.
A model reasons. An agent also has a place to work, context to return to and permission to act. Your product decides when it should.
The agent opens an approved source, prepares a brief and saves the first version in its own files.
Illustrative lifecycle, not live telemetry or a cost estimate. Wakes, tools and actions stay within the permissions you set.
Your agent is more than the model behind it. Its personality, files, skills and permissions belong to your product, not to a chat session.
A frontier model for complex reasoning and code. Model access is separate from your agent's identity and home.
Model selection illustration. Availability depends on runtime compatibility, provider access and credentials. Selecting here does not change a live agent.
Delegate work, request wakes and inspect status with scoped service keys. Keep your customer experience in your own product.
curl -sS -X POST "$ACTA_BASE_URL/v1/principals/$PRINCIPAL_ID/delegations" \
-H "x-acta-key: $ACTA_SERVICE_KEY" \
-H "idempotency-key: msg_123" \
-H "content-type: application/json" \
--data '{"source":"api","sourceMessageId":"msg_123","text":"Research this account and prepare a brief."}'Authenticate your backend with scoped credentials.
Use a customer identity to provision a runtime and delegate tasks.
Inspect health, results and recorded usage in the console.
View runtime health, delegation results, stored agent state and recorded usage in your workspace.
Open your consoleSelf-service billing is not yet available. Usage visibility is available in the console.
Software companies building products where each customer needs an agent with continuing memory, files and tools. Your company integrates ACTA; your customers stay in your product.
Your customer experience, agent instructions, skills, policies, approved integrations and model configuration. ACTA operates the runtime and keeps it associated with the customer's persistent agent identity.
No. Work can begin with an API request, a message, a scheduled wake or a product event. Proactive work needs explicit permissions and bounds; it is not unrestricted background activity.
No. Persistent identity and state remain while compute sleeps. ACTA can wake the runtime for approved work. Around-the-clock availability is a lifecycle design, not a guarantee of continuous computation or an uptime SLA.
Model configuration is separate from persistent agent identity. The current native-Hermes pilot uses GPT-5 mini through OpenRouter. Other models shown are model examples, subject to provider access and integration validation, not a claim that every model is enabled.
No. It is an interactive lifecycle illustration. Sign in to the console to inspect your workspace's actual agents, runtime state and recorded activity.