What is Ashcroft AI?
Ashcroft AI is a managed agent operations company. We help teams implement, govern, and maintain AI-agent workflows — concierge-led setup and ongoing support — so operational knowledge doesn't leave when your people do. Read-only first, with humans in control.
Concierge first, maintained over time
Most companies aren't deep enough into AI to implement and maintain agents themselves — and when an employee leaves, the workflow leaves with them. An Ashcroft team walks you through implementing your agent workflows, governs them with human approval, and maintains them over time as durable, company-owned infrastructure. The platform below is the operating surface; the service is the product. A self-serve version is a future plan.
Capture, memory, governance, routing, runtime
- Capture. Pull in notes, threads, requests, and documents without forcing structure up front.
- Operational memory. Keep a durable, governed record of work and context instead of scattered silos.
- Governance. Every state change is policy-checked and recorded in an append-only trail.
- Routing. Send the right work to the right people, queues, or agents — by rule, not guesswork.
- Agent runtime. Agents run with tool permissions, risk rules, sandboxing, and audit.
Read-only first & human approval gates
Automation is a privilege, not a default. Surfaces read data; writes and external dispatch each require an explicit, named gate. Today the platform runs read-only: two read-only agents are active with three internal read-only tools and zero external, write, or dispatch capability. Nothing external is connected.
Connection ownership model
Connections belong to four distinct owners so it is always clear what a credential does: User Identity (how you sign in), User Data (your own mailbox/calendar, a separate consented step), Platform Provider Keys (server-only AI keys that never sign you in), and Agent Tool Grants (which agent may use which capability). Adding a provider key never auto-grants it to an agent.
Frequently asked questions
- What is Ashcroft AI?
- Ashcroft AI is a managed agent operations company and its platform — the private office for your operations; Ash is the concierge. We help teams implement, govern, and maintain AI-agent workflows so operational knowledge does not disappear when employees leave. The Ash app is the operating surface for that service — read-only first, with humans in control.
- Is this a self-serve app or a service?
- Today it is a concierge service: an Ashcroft team personally onboards each client — workflow audit, mapping, agent design, governance rules, and integration walkthroughs — then maintains the workflows over time for a recurring fee. A fast, mass-scale self-serve product is a future plan, not the near-term offer.
- What does "workflow continuity" mean?
- Your workflows should not leave when your people do. We capture each workflow and its context as durable, company-owned infrastructure, and tie agent permissions to the company rather than to a person — so when an employee transitions out, the workflow keeps running and the knowledge is preserved.
- What does "read-only first" mean?
- The application reads and organizes data by default. Any write, external dispatch, or integration requires an explicit, named human approval gate. The system fails closed: it does nothing on the outside world until authorized.
- How does the agent runtime stay safe?
- Agents run inside a control plane with explicit tool permissions, risk rules, sandbox testing, and an append-only audit trail. Today two read-only agents are active with three internal read-only tools and zero external, write, or dispatch capability. Any tool can be revoked and any agent deactivated at any time.
- What are human approval gates?
- Every consequential action waits for an explicit authorization gate before it runs. Gates cover things like connecting an integration, granting an agent a tool, writing data, or deploying. Nothing consequential happens automatically.
- What is the connection ownership model?
- Connections belong to one of four owners: User Identity (how you sign in), User Data (your own mailbox/calendar, a separate consented step), Platform Provider Keys (server-only AI keys that never sign you in), and Agent Tool Grants (which agent may use which capability). Adding a provider key never auto-grants it to an agent.
- Is anything connected to external services?
- No. As of this writing no external provider is connected, no OAuth has started, and nothing is deployed. The integration surfaces are staged and ready for credentials, and each connection requires its own gate.
See it for yourself
Step into Ash, or open the connection setup to see exactly what each integration needs before anything goes live.