onesquad.ai
April 1, 2026
New Demo: Privacy & Deletion Audit
#
See @os-privacy audit every data store for deletion completeness — and find where the cascade breaks at the infrastructure boundary.
What the demo covers
Deletion cascade audit
- Architecture & data model read — @os-privacy reads the architecture, data model, and deletion handler to map every store that holds user data.
- Application layer: 4/4 pass — Aurora (CASCADE DELETE), DynamoDB (explicit delete), S3 uploads (prefix delete), and Cognito (admin API) all have working deletion paths.
- Infrastructure layer: 0/3 pass — S3 exports bucket, CloudWatch Logs (2-year retention), and CloudWatch Metrics (immutable dimensions) are all unreached by the deletion cascade.
- Remediation & artifacts — Prioritized P0–P2 remediation per failing store, plus
data-inventory.md (full audit table) and data-flow-map.md (Mermaid diagram showing where deletion stops).
Roles featured
@os-privacy
v0.1.2 Beta
March 31, 2026
New Workflow Types, Two New AI Roles & Role Improvements
#
Five new predefined workflow types, two new AI roles, and a set of role improvements that make the whole squad sharper.
What's New
Additional Workflow Types
Five new predefined workflow tools join the existing Feature workflow:
fix_bug — diagnose, fix, verify.
new_product — discover, architect.
rca — root cause analysis.
cost_analysis — FinOps cost and ROI analysis.
architecture_review — architecture impact assessment.
Two New AI Roles
@os-analytics — Analytics Engineer. Owns the measurement layer: writes event taxonomies, pipeline specs, and dashboard requirements. Hands off to @os-dev for instrumentation and @os-devops for pipeline provisioning.
@os-finops — FinOps Advisor. Reviews architecture for cost exposure before and after launch, proposes changes to dev and staging environments without touching production, and tracks cost ROI against accepted recommendations.
AI Role Improvements
- Role invocation is now the canonical activity signal, replacing the previous started/completed event pair.
- Role instructions are now technology-agnostic — no specific stacks are referenced in role prompts.
- Build agents must read and conform to the coding plan's API Contract section before generating code.
- Test agents derive expectations from the spec, not from handler output.
Known Limitations
- Self-service billing not yet available — tier changes are admin-only.
- Email notifications for invitations not yet implemented.
- Organization deletion and ownership transfer not supported.
What's Next
- Self-service billing integration (Stripe).
- Email notifications for org invitations.
- Enterprise tier (SSO, unlimited seats).
onesquad.ai
March 30, 2026
New Demo: Analytics Dashboard
#
See the full analytics workflow in action — from a raw operator request to a live interactive dashboard, with @os-analytics, @os-architect, and @os-ux all collaborating across 4 phases.
What the demo covers
4-phase walkthrough
- Phase 1 — Analytics Discovery — @os-analytics gathers latency, infra, and scale requirements, then proposes a DynamoDB Streams → S3 micro-batch → Iceberg → Athena pipeline and writes
pipeline-arch.md.
- Phase 2 — Architecture Review — @os-architect validates the new datalake stack against the existing system: IAM boundaries, Lambda fan-out, and Glue catalog cost. Operator approves.
- Phase 3 — Requirements — @os-analytics writes
requirements.md with full event taxonomy, three dashboard specs, Athena query list, and shared filters.
- Phase 4 — Dashboard Wireframe — @os-ux builds an interactive 3-view wireframe: Activation Funnel, Role Usage Heatmap, and Retention Cohorts. Operator approves.
Roles featured
@os-analytics · @os-architect · @os-ux
onesquad.ai
March 30, 2026
New Role: @os-analytics
#
The Analytics Engineer joins the squad. @os-analytics owns the "what to measure" layer — sitting between product decisions and the code that instruments them.
What @os-analytics does
Three workflows
- Product setup — For a new product with no analytics yet: produces
metrics.md, event-taxonomy.md, pipeline-design.md, schema.md, and dashboard-spec.md.
- Feature review — When a new feature spec is written: updates event taxonomy, metrics, and dashboard spec to cover the feature's events.
- Validation — Before ship: verifies events fire correctly, queries return expected data, and dashboards render.
Handoffs
@os-analytics defines the events → @os-dev instruments them. Designs the pipeline → @os-devops provisions it. Specs the dashboard panels → @os-dev builds them. Reviews feature specs for analytics gaps → @os-pm updates the spec.
onesquad.ai
March 28, 2026
New Demos: UX & Wireframe and FinOps Review
#
Two new interactive demos join the gallery — @os-ux going from feature spec to annotated wireframe, and @os-finops recovering $558 in pre-launch infrastructure costs before a single user signs up.
What's in the demos
- UX & Wireframe — @os-ux reads the feature spec, maps the user journey, identifies key screens and edge cases, and produces an interactive wireframe with annotated component breakdowns. The session ends with a structured design handoff summary ready for @os-dev to consume.
- FinOps Review — "The $558 You Haven't Spent Yet" — @os-finops reads the architecture before launch, spots that dev and staging are provisioned identically to production — App Runner and Aurora running always-on, serving zero users — and proposes two CDK property changes (
minInstances: 0, minCapacity: 0) that recover $558 in the first 6 months. Production: untouched.
v0.1.1 Beta
March 28, 2026
Multi-User Organizations & Role-Based Access
#
Multi-user organizations, role-based pricing, and a major database consolidation. OneSquad is no longer single-user.
What's New
Organizations & Team Invitations
Users can now belong to multiple organizations. Organization admins invite members by GitHub username and manage roles, API keys, and workspaces — all fully isolated per org. Pending invitations are shown at login and can be accepted or declined. Three plan tiers: Free, Professional, and Organization (3 seats).
Role-Based Freemium Gate
Spec counting is replaced by role access. Free users get @os-pm and @os-dev. Professional and Organization users unlock all 7 roles — PM, TPM, Architect, UX, Dev, DevOps, and SDET.
Invitation Code Gate
Admins can require a one-time invitation code to access OneSquad. Codes are generated in bulk and can be revoked at any time. Users with a pending org invitation bypass the gate automatically.
MCP Rate Limiting
Per-user hourly rate limits on MCP requests, defaulting to 60 per hour. Admin-configurable per org. Returns 429 with minutes-until-reset when exceeded.
AI Role Improvements
get_role now accepts a context parameter for loading role-specific procedures (e.g., get_role('dev', 'build')). Specs, plans, and test directories are configurable via config.md. Standardized handoff messages between roles. @os-sdet includes contract tests; @os-dev and @os-tpm enforce API contracts at plan gate. The MCP endpoint is now simply /v1/mcp.
Known Limitations
- Self-service billing not yet available — tier changes are admin-only
- Email notifications for invitations not yet implemented
- Invitations require exact GitHub username — no search
- Ownership transfer and organization deletion not supported
What's Next
- Self-service billing integration (Stripe)
- Email notifications for org invitations
- Enterprise tier (SSO, unlimited seats, custom personas)
onesquad.ai
March 26, 2026
A new demos page on onesquad.ai brings all three interactive walkthroughs together in one place — with screenshots and direct launch links.
What's on the page
- Bootstrap OneSquad — a single MCP command provisions the
.onesquad/ workspace from the server, then @os-pm leads Product Discovery and @os-architect drives System Design before a line of code is written.
- Ship a Feature — a ticket becomes a spec, plan, code, tests, and production deploy.
- Bug Triage & RCA — @os-tpm coordinates the triage workflow, dispatching each role as a sub-agent. @os-sdet reproduces the incident and narrows the blast radius; @os-dev produces a structured RCA doc (failure mode, contributing factors, fix rationale); @os-devops confirms the fix is deployed and regression-proof.
v0.1.0 Beta
March 22, 2026
OneSquad is a spec-driven SDLC for AI-powered teams. PMs write specs. Agents build from specs. Nothing bypasses the gate. v0.1.0 is the foundation.
What's New
MCP AI Roles
OneSquad AI Roles — @os-pm, @os-architect, @os-dev, @os-ux, @os-tpm, @os-devops, @os-sdet — are served from a centralised MCP server over streamable HTTP. Persona instructions live on the server, never on the user's machine, and update without any client-side action. Connect your IDE with an API key and a workspace ID and your Roles are live.
A bootstrap MCP tool scaffolds the full .onesquad/ directory in any project with a single command — no zip download, no manual file copying.
Workspaces & API Keys
The unit of work in OneSquad is a workspace. Create one from the dashboard to get a workspace ID, generate an API key (shown once, stored as a hash), and paste both into your IDE MCP config. The MCP endpoint follows the pattern /v1/workspaces/:workspaceId/mcp.
Freemium — Free Plan
All users start on the Free plan: 10 spec generations per month, tracked across both the web app and MCP. The monthly counter and remaining quota are visible in the top nav.
Known Limitations
- Only the Free plan is enforced. Pro, Team, and Enterprise tiers are defined in the data model but not yet billed.
- Organisations are single-user. Invitations are not available yet.
What's Next
- Billing integration for paid tiers
- Multi-seat org invitations