Research Orchestration for Human Researchers and AI Agents
中文
Synapse is a research orchestration platform that brings human researchers and AI agents together. It manages the full research lifecycle — from literature review and question formulation through experiment execution and report generation — with built-in agent management, compute orchestration, and real-time observability.
Inspired by the AI-DLC (AI-Driven Development Lifecycle) methodology and built upon Chorus.
Milestone: We reproduced Karpathy's auto-research task and outcome pattern on Synapse. Unlike designing the behavior of a single agent through a program.md, Synapse treats autonomous experimentation as a native platform capability. Research projects, experiment lifecycles, compute allocation, and context accumulation are all managed through structured MCP tools, creating dynamic contexting. Research memory is persisted in the platform and exposed as tool-based state that agents can call, fetch, and expand on demand according to the current situation, rather than forcing everything to be loaded into an agent's context window up front.
What's New
v0.6.1 — Experiment Board UI Polish (2026-04-15) 🔴 New
- Project descriptions on the dashboard now preserve intentional blank lines when expanded, improving readability for structured briefs
- The experiment plan side panel now renders above the detail sheet overlay, uses the shared panel background, and allows normal text selection
- The autonomous experiment entry button is more prominent, with refreshed blue styling and a yellow lightning icon for faster scanning
- The experiment detail sheet is wider on large screens, giving experiment plans, results, and progress logs more room
v0.6.0 — Agent Types & Research Copilot (2026-04-12)
- Agents now have a
type field (OpenClaw or Claude Code) with internal transport mapping — Web UI dispatch features only show realtime-capable agents
- Claude Code Research Copilot: SessionStart presents projects with progress summaries and guides users through the research lifecycle (paper search → deep research → questions → experiments → analysis)
synapse_checkin returns assigned experiments and project progress for intelligent workflow suggestions- Agent management UI gains a type selector and badge display
v0.5.1 — DeepXiv Integration (2026-04-10)
- Paper search now uses DeepXiv hybrid search (BM25 + vector) over arXiv, with arXiv API as fallback. Removed Semantic Scholar and OpenAlex.
- Agents can read full paper content via progressive reading tools:
synapse_read_paper_brief, synapse_read_paper_head, synapse_read_paper_section, synapse_read_paper_full
- Deep research literature reviews are now based on actual paper content, not just abstracts
- DeepXiv token configurable from Settings > Integrations
v0.5.0 — Autonomous Loop & Related Works (2026-03-29)
- Autonomous experiment loop: agents propose → humans review → agents execute
- Related Works page with auto-search, manual arXiv URL addition, and deep research reports
- Experiment live status tracking (sent/ack/checking/queuing/running)
- Compute pool binding per project
Table of Contents
Vibe Research
Vibe Coding showed that people can describe intent and let AI handle execution. Vibe Research applies that same shift to the research lifecycle:
Humans set direction. Agents execute, report, propose, and iterate. Humans review, steer, and decide.
Stages of Agent Autonomy in Research
Synapse is built to move research teams through these stages deliberately.