---
title: "Technical Specs | Achiral AI Contextual Intelligence"
description: "Technical specifications for Achiral AI: shared organizational memory, Chiro, personal executive assistants, tenant isolation, connectors, private inference, and security controls."
---

# Achiral Technical Specs

Canonical HTML: https://achiral.ai/specs

Achiral is an Emergent ACT-R Memory System for teams and AI agents.

## Direct Answer

Achiral turns team activity into private, persistent context. Chiro and personal executive assistants retrieve organic memory, understand workflow state, and act through official tools so teams and agents do not have to rebuild context from scratch.

## Architecture

| Layer | Specification |
| --- | --- |
| Canonical product | ACT-R-inspired memory layer for AI |
| Org assistant | Chiro, an org-wide assistant backed by the organization memory tenant |
| Personal assistants | Executive assistants per individual, with private memory and preferences |
| Memory model | Tenant-scoped vector memory plus operational state |
| Retrieval | Hybrid semantic and keyword retrieval with memory-aware ranking |
| Connectors | Slack, GitHub, HubSpot, Zendesk, Intercom, Notion, Asana, Linear, Jira, Google, Microsoft 365, Salesforce, and other operational systems |
| Actions | Audited action execution with policy checks, approval gates, tool permissions, and resumable workflows |

## Platform Capabilities

The full surface of Achiral's runtime: memory, assistants, retrieval, connectors, action execution, and controls.

- Org-wide Chiro
- Personal EAs
- Shared memory
- Private user memory
- Hybrid retrieval
- ACT-R memory scoring
- Connectors
- MCP servers
- Actions
- Team approvals
- Audit trails
- Tenant isolation
- Private inference
- Workflow intelligence
- Human guidance

## Memory Layers

- Episodic memory: recent conversations, decisions, events, tasks, and workflow context.
- Semantic memory: durable facts, recurring workflows, preferences, policies, customer context, and operating patterns.
- Core memory: human-validated facts that should rarely decay.
- Reinforcement: useful memories gain weight when retrieved and reused.
- Decay: low-signal context can lose influence or require review before reuse.
- Human validation: admins and Shepherds can guide memory quality.

## Security Posture

- Tenant isolation by organization, user, assistant, role, and connector permissions.
- TLS in transit and encrypted storage for credentials, memory, connector tokens, and operational records.
- Role-aware retrieval, approval policies, audit logs, onboarding, and organization-scoped permissions.
- Customer data is not pooled across organizations and is not used to train a shared customer model.
- SOC 2 audit work is in progress.

## Infrastructure and Integrations

| Component | Detail |
| --- | --- |
| Runtime | Next.js web app, API services, worker queues, websocket chat, and Kubernetes-managed production deployments. |
| Storage | MongoDB for operational state, Weaviate for tenant-scoped vector memory, Redis for queues/rate limits/leader election, and object storage for uploads. |
| Inference | Private inference-first model routing with escalation paths for complex tasks when configured by the organization. |
| Connectors | Slack, GitHub, HubSpot, Zendesk, Intercom, Notion, Asana, Linear, Jira, Google, Microsoft 365, Salesforce, and other operational systems. |
| MCP | Achiral can expose org context to external agents and register external MCP servers as live context/action sources. |
| Actions | Audited action execution with policy checks, approval gates, tool permissions, and resumable workflows. |

## Memory-Bucket Pricing

Pricing is bucketed by team size and memory capacity. Achiral avoids per-token surprise billing because the value is continuity of context across the team, not raw token volume.

| Tier | Price | Team size | Included context |
| --- | --- | --- | --- |
| Founder | $0/mo | 1-2 seats | Org-wide Chiro plus personal EA basics for small teams starting out. |
| Spark | $30/seat/mo | 3-10 seats | Shared org memory, personal EAs, connector setup, and predictable team pricing. |
| Seed | $29/seat/mo | 11-25 seats | Larger memory rate band, priority support, custom knowledge base, and deeper onboarding. |
| Rise | $28/seat/mo | 26-50 seats | Expanded team capacity, stronger workflow coverage, and higher operating context. |
| Boost | $27/seat/mo | 51-100 seats | Expanding teams with broader workflow coverage. |
| Scale | $26/seat/mo | 101-250 seats | Large memory rate band, isolated infrastructure options, higher concurrency, and enterprise support. |
| Grow | $24/seat/mo | 251-500 seats | Larger organizations before custom enterprise terms. |

## Honest Constraints

- Achiral is memory-first. Teams looking only for generic internet chat may not need the platform.
- The strongest results require connected tools, team adoption, and enough operational history for memory to be useful.
- SOC 2 audit work is in progress; Achiral should not be described as SOC 2 certified until the audit is complete.
- Production runs on infrastructure Achiral operates, shared across organizations, with each organization's memory isolated in its own tenant.

Need to validate your architecture fit? Achiral can map Chiro, memory, connectors, and permissions to your current operating model.

## Related

- [Security](https://achiral.ai/security)
- [Pricing](https://achiral.ai/pricing)
- [Compare Achiral](https://achiral.ai/compare)
- [ACT-R memory architecture](https://achiral.ai/concepts/act-r-memory-architecture)
