Choose by bottleneck

Start with the problem you already feel.

The Agent Series is intentionally split into composable projects. This page explains which planned OSS release maps to each repeated bottleneck; public repository and package links will appear when the individual releases are ready.

Quick decision

If your main problem is… Planned OSS project Role in the stack
Multiple coding agents repeat work or lose task context Agent Workbench Operate multi-agent tasks, ownership, handoffs, review and memory
Agents/editors read too much repository context Agent Token Optimizer Audit and optimize repository context before execution
AI review misses user/adopter/operator perspectives Agent Persona Add reusable human-perspective review lenses
You are researching explicit agent continuity Agent Human Memory Explicit memory, authority, provenance and forgetting
Reusable agent skills need packaging across tools Agent Skill Shelf Canonical skill source and multi-agent outputs
One output needs multi-perspective synthesis/judging Fusion Skill Independent reviewer lanes and judge synthesis
Product ideas need briefs, PRDs and task plans AI Product Manager Product planning and structured implementation artifacts
Tasks need deterministic assignment before execution Agent Task Router Route work to agents, model tiers, skills or human gates
Structured tasks need target-specific handoff packets Agent Handoff Kit Render agent-ready handoff packets
Agent activity needs reports, risks and next actions Agent Reporting Explain runs, failures, learning and next actions
Developer-machine performance is slowing agent work Agent Machine Doctor Diagnose machine health and developer-tool performance
Repository / agent security needs a release gate Agent Security Doctor Security diagnostics and release gating
Human approvals/results need Slack or Discord delivery Agent Channel Bridge Human-in-the-loop external channel bridge

See descriptions and release status for all 13 projects →

How the anchor projects fit together

Coordinate → Agent Workbench

The operating layer: tasks, routing, claim leases, prompts, handoffs, reports, review, and completion memory.

Optimize → Token Optimizer

The repository preflight: context waste, instruction bloat, editor/indexing risk, token budgets, and routing signals.

Review → Agent Persona

The human-perspective review layer: reusable adopter, user, security, operator, buyer, and contributor viewpoints.

Remember → Human Memory

The experimental memory layer: explicit memory with provenance, authority, review, retention, and forgetting.

Public release policy

All 13 projects are planned for public OSS release. Until each repository or package is actually accessible, the public site shows product direction and synthetic previews rather than private links, unpublished install commands, or public recruitment flows.

Portfolio rule: the complete roadmap is public, but availability is not implied. Release links are added only when each project clears its packaging, documentation, licensing, and safety gates.