Zero cold start. No duplicate work. No wasted AI spend.
Open source • Bring your own keys • Limited pilot spots
✅ Session complete: retention pain points mapped
Added to team knowledge base:
→ 5 insights on SMB churn triggers
→ 2 product decisions: onboarding over activation
→ Interview framework filed under /customers/smb/
Retrieved from team knowledge base:
→ Tom's pricing research (847 tokens — saved ~6,800)
→ Jordan's onboarding drop-off analysis (623 tokens — saved ~5,200)
Quota: ████████░░ 68% used • resets in 4h 12m
Background task: PRD-47 gap analysis queued ↗
Read every line. The MCP server, CLI, and hooks are fully open. No black boxes handling your team's knowledge.
View on GitHubLLM calls go directly from your machine to your provider. Evols never proxies or stores your inference traffic.
Run Evols on your own infra. Your team's knowledge graph never leaves your environment if you choose not to.
Self-host docsAI tools are built for individuals. Teams using them collectively suffer compounding failures that no single tool today addresses.
PM research, engineering decisions, UX insights — every teammate rebuilds from scratch. The same 12,000 tokens of context compiled daily across every role.
Two engineers solve the same infra problem. Three PMs research the same competitor. Work collides at review — after the tokens are burned.
Some teammates hit session lockouts mid-task. Others let quota reset unused. No tool shows collective AI capacity across the team.
AI drafts look complete. Teammates spend time verifying, rewriting, clarifying — the hidden rework tax on every AI-assisted deliverable.
Agentic systems span models, orchestrators, APIs, logs, dashboards. When something breaks, engineers reconstruct what happened from scratch.
Teams underestimate how fast inference costs scale. Productivity gains get erased by optimization work and unplanned budget pressure.
No manual configuration. Evols configures hooks, MCP server, skills and team context.
$ curl -fsSL https://api.evols.ai/api/v1/install/script | sh
$ evols login
$ evols install
✓ Hooks registered • MCP server active • Team context loaded
✓ Done. Your team's knowledge base is ready.
Evols fires the SessionStart hook. Team knowledge base context is injected immediately. No cold start. No accumulation period. Day one.
UserPromptSubmit hook checks for redundant work. Injects relevant prior context. Tracks session token usage client-side in real time.
Stop hook syncs outcomes to the team knowledge base. StopFailure hook detects rate limits and queues work for the next window.
New teammates inherit the team knowledge graph on day one. No setup, no accumulation period — full context from the first session.
Every AI session contributes to a shared team knowledge base automatically. Query it instead of rebuilding it — at 8× fewer tokens.
Detect duplicate AI work before tokens are burned — not after the work collides at code review or the decks land in the same meeting.
See collective team AI capacity in real time. Redirect expiring quota to backlog tasks before it resets unused.
Project-level and role-level context permissions. Teams control who sees what knowledge — built for orgs that handle sensitive context.
One CLI install activates everything. MCP server and hooks are wired automatically into Claude Code, Cursor, Zed, Codex, and Antigravity.
Evols is a team AI brain that gives every member shared context across Claude Code, Cursor, Zed, and other AI coding tools. It eliminates duplicate work, provides zero cold start for new teammates, and shows collective AI quota usage in real time.
Evols stores AI session outputs in a shared knowledge graph. Instead of each teammate compiling the same context from scratch, they retrieve it at approximately 8× fewer tokens per session — cutting redundant usage across the whole team.
Evols works with Claude Code, Cursor, Zed, GitHub Copilot (Kiro), Cline, and Codex via a single CLI install that registers MCP server hooks automatically.
The handoff tax is the wasted time when context built in one AI session cannot transfer to the next person. A developer's findings, decisions, and reasoning disappear when the session closes — forcing every teammate to start from scratch and compile the same tokens over and over.
Run: curl -fsSL https://api.evols.ai/api/v1/install/script | sh, then evols login and evols install. Hooks and MCP server are registered automatically into Claude Code, Cursor, and Zed.
"I think there is room here for an incredible new product instead of a hacky collection of scripts."
"We're stopping at individual productivity. One department doesn't know what the other one does. That's where it cracks."
Token exhaustion documented by BBC, Forbes, MacRumors — March 2026
VCs calling team context infrastructure "the next big thing in AI" — Forbes, April 2026
GitHub Copilot Workspace shut down. No replacement exists.
Sessions should compound, not disappear.
Active with Claude Code, Copilot, Kiro, Cursor, and Cline