A personal AI brain for context-aware agent teams.
OpenAnimus helps humans and AI agent teams turn trusted context into reviewed action with better memory, evidence, and control.
North star
personal AI brain
First wedge
agent-team operations
Default posture
trusted context + human review
Personal brain
Risk: medium
Issue #12
Auth flow changed
Prior PR
Local-first default
Decision note
Agent operations
Routed workflow
Send this to frontend-dev + QA, then require evidence before approval.
Decision boundary
Agents prepare work and evidence. People approve important outcomes.
The problem
The issue, PR, memo, or decision is only the visible artifact. The real work lives in context, prior decisions, run history, review comments, and the evidence behind every change.
Personal context is scattered across tools, conversations, files, issues, and project history.
Coding agents can act, but they often start without the task risk, prior decisions, or evidence expectations.
Human review becomes expensive when the run state, changed files, tests, screenshots, and unresolved risks are buried in logs.
How the agent loop works
OpenAnimus gives the personal brain a concrete job: route agent work with memory, risk, evidence, and human approval.
Bring together the project context you choose to trust: issues, notes, files, PRs, decisions, calls, and optional capture systems.
Turn scattered work into linked decisions, summaries, open questions, risks, and source-backed memory.
Give each agent the right context pack, task risk, workflow path, and expected evidence before it acts.
Keep approvals with people by surfacing changed files, commands, tests, screenshots, PRs, and unresolved risks.
Wedges
OpenAnimus starts with software builders because the loop is concrete and reviewable. The same pattern can later expand into broader personal and business context.
Builders
OpenAnimus starts where the loop is concrete: issues become agent workflows, evidence, PRs, and human-reviewed decisions.
Later: Work and life
The same pattern can expand carefully into richer personal context while keeping capture intentional and judgment human.
Architecture
OpenAnimus should integrate pragmatic context sources and agent runtimes while owning memory, workflow routing, evidence, and review.
L1
Files, notes, issues, PRs, decisions, calls, tools, and optional screen or browser context.
L2
Memory, summaries, entities, provenance, retrieval, open questions, and decision history.
L3
Task risk, workflow routing, context packs, run state, blockers, evidence, and PRs.
L4
Dashboards, context views, evidence trails, approvals, stream mode, and review controls.
Principles
OpenAnimus is designed for trusted context and reviewed action, so the product boundary matters.
OpenAnimus should work with context you choose to bring in, with clear control over what is captured, stored, shared, or sent to models.
Agents should stage context, changed files, commands, tests, screenshots, risks, and review notes before important work is approved.
OpenAnimus can help prepare and route work, but it should not claim to replace expert judgment or make high-stakes decisions autonomously.
Open source
Open issues, challenge the architecture, and describe the agent workflows where trusted context and human review would change the outcome.
Built in public, product first
The project is developed openly so the architecture, tradeoffs, and agent workflows can be inspected. The stream is a proof of process, not the product.
Community discussion and workflow feedback.
Short product updates and build notes.
Short answers for the current product direction.
Long-term, OpenAnimus is a local-first personal AI brain. The first wedge is narrower: agent-team operations for software builders.
No. OpenAnimus is designed to stage trusted context, evidence, drafts, and review trails so people can make better decisions.
No. OpenAnimus should work with context you choose to bring in. Capture and data sharing should be intentional choices, not always-on defaults.
Software builders using AI agents on real project context: issues, decisions, files, PRs, tests, run history, and reviews.