Private AI Scrum Master for engineering teams

Run better standups without sending your team's voice or delivery data to the cloud.

Scrum Master AI facilitates daily Scrum conversations, captures developer updates, detects blockers, creates minutes, and prepares stakeholder summaries from a local-first workspace with policy-controlled model routing and a tested agent runtime.

Request private pilot See how it works

Daily Scrum control center

Hi Asha. What did you complete since the last stand-up?

Finished login validation and started story API testing.

Captured. Any blockers?

CompletedLogin validation

TodayStory API testing

BlockersAwaiting update

Local voiceRaw audio discarded after transcription

50 golden testsConversation benchmark for no-repeat and completion behavior

Project-awareStories, roles, blockers, actions, minutes

Provider-neutralLM Studio, vLLM, Bedrock, compatible endpoints

Product

A Scrum Master that actually runs the meeting.

Teams do not need another form to fill. They need a facilitator that asks the right follow-up, remembers what was answered, links updates to work, and turns the conversation into useful delivery records.

Conversational standupsHands-free voice flow with local speech-to-text, meeting WebSockets, local LLM reasoning, and local text-to-speech.

Structured updatesExtracts completed work, planned work, blockers, story references, risks, and actions with deterministic guardrails.

Rally-style workspaceProjects, teams, stories, meetings, minutes, actions, and delivery dashboards in one place.

Tool runtime32 Scrum tool contracts for stories, blockers, actions, risks, decisions, assignments, minutes, and escalation.

AI Gateway

Local by default. Cloud only when policy allows it.

The application now uses a provider-neutral model layer. Teams can keep conversational facilitation on LM Studio or vLLM, use a hardened local LoRA SLM for Scrum fact extraction, or enable Amazon Bedrock for approved workloads without rewriting the Scrum Master agent.

Local Scrum SLMFine-tuned adapter for yesterday, today, blockers, story references, and incomplete updates.

LM StudioDeveloper and small-team local runtime for private pilots.

vLLMRecommended local production runtime for higher throughput.

Amazon BedrockNative Converse API path with IAM, streaming, request metadata, and guardrail support.

Workflow

From meeting to minutes in one flow.

1Create project

Add team members with roles like developer, tester, project manager, product owner, and stakeholder.

2Start standup

The AI facilitator asks for updates, listens locally, and avoids repeating already answered questions.

3Review minutes

Scrum Master reviews captured work, blockers, risks, decisions, and actions before publishing.

4Share summary

Stakeholder email is previewed first and sent only after explicit confirmation.

8.6 / 10Current internal pilot readiness rating

50 / 50Golden conversation scenarios passing

32 toolsValidated Scrum tool contracts

10 usersSimulated standup smoke test passing

Demo flow

Show the buyer a real Scrum Master workflow.

The pilot demo walks through admin controls, project roles, backlog context, a hands-free standup, tool-backed story creation, and approved minutes. It is designed to prove that the agent can facilitate, remember, act, and finish.

1. Ask about assigned stories The agent reads project context before continuing the standup.

2. Give yesterday/today/blockers The agent records structured updates without repeating answered questions.

3. Create a story naturally The model selects the tool; deterministic policy validates the action.

4. Finish with minutes The meeting closes only after the ceremony update is complete.

Security and privacy

Built for teams that cannot casually leak delivery data.

The pilot version is designed for private evaluation. The application runs with MySQL, local AI by default, local voice processing, signed login sessions, role-based permissions, model-route governance, and audited material actions.

Raw audio is deleted immediately after transcription.
Transcript retention jobs are scheduled for 90 days.
Cloud inference is disabled unless an administrator enables explicit route policy.
External summaries require human approval before send.
Golden benchmark covers completion, no-repeat behavior, story lookup, story creation, and done intent.
Admin console shows AI providers, model routes, policies, integrations, outbox, audit activity, and operations metrics.
Enterprise roadmap includes OIDC, Jira, Google Workspace, workers, and HA deployment.

Private pilot

Start with one engineering team.

AI Scrum Master Private Pilot

For teams that want to test local AI facilitation before committing to a full enterprise rollout.

Up to 10 users One project team Local or governed cloud AI Daily standups Minutes and actions Stakeholder summaries

Contact

Request a private pilot.

Use this first website to collect interested teams. Replace the email below with your business email when ready.