Work

Momental gives you four specialized agents that handle the full lifecycle of product work — from planning through shipping. Assign them to tasks just like you'd assign a team member, from the task panel, the Plans page, or via the API.

Building your strategy tree

Start with your objectives — what does your team need to achieve this quarter? Momental helps you decompose them step by step:

Vision → Objective → Key Result → Solution → Epic → Task

You don't have to build the tree manually. Momental can generate the next level for you:

The tree is always live. Completed work rolls up automatically. Blocked items surface. Nothing hides.


Agents

Planner

The Planner decomposes complex goals into concrete, actionable work. Give it an objective and it returns a structured plan: epics, tasks, acceptance criteria, suggested sequencing. It understands your team's strategy context, so plans reflect your actual priorities — not a generic template.

Best for: breaking down solutions, quarterly planning, creating roadmaps from scratch.


Researcher

The Researcher digs into your knowledge graph and the web to surface what your team already knows — and what it should. It searches atoms, scans documents, queries external sources, and synthesizes findings into clear, cited summaries that become atoms in your graph.

Best for: competitive research, pre-build context gathering, answering "what do we already know about X?"


Writer

The Writer produces polished content grounded in your team's actual context. Blog posts, documentation, release notes, proposals — all informed by your knowledge graph, your product tree, and your brand voice. No hallucinating features that don't exist.

Best for: documentation, marketing content, internal memos, changelog entries.


Coder

The Coder is your autonomous coding agent. Assign it a task in Momental, and it spins up an isolated cloud container, clones your repository, understands the codebase semantically, implements the feature or fix, runs your tests, and opens a pull request. It never stores your source code. Each container is destroyed after the task completes.

Works best on tasks with clear acceptance criteria — the more specific, the better.

Best for: feature implementation, bug fixes, test writing, refactoring, documentation generation.


Vega — Code Guardian

Vega reviews every pull request automatically — before a human ever looks at it. It checks for security vulnerabilities, flags anti-patterns, analyzes the blast radius of changed functions, and enforces your team's engineering principles stored in the knowledge graph.

Risk levelWhat it means
LOWNo auth, database, or API contract changes → safe to auto-merge
MEDIUMLogic changes with potential edge cases → worth a human look
HIGHAuth, payments, migrations, breaking changes → requires review

Vega posts its review as a GitHub comment within seconds of a PR being opened. Zero configuration required to start.

Install Vega: Subscribe to Vega, go to Settings → Connections → GitHub, and install the Vega GitHub App. Every new PR is reviewed automatically from that point on. Optionally add a .vega-agent.yml to your repo root to customize rules.