BETA In open beta. Install live. Lock $5/mo for your first 12 months. See pricing →
Claude Code · VS Code Copilot · Codex CLI · Mistral Vibe · MCP

Stop re-explaining your codebase to AI.

Captures the decisions you just made. Injects them back before your next prompt. Automatically, locally, free forever. 

curl -fsSL gethydrate.dev/install | sh
macOS Linux Windows Works in 30 seconds

Local-first. Native Claude Code hooks. Hydration Packs. Team memory via git (early access). Extractive, no LLM tokens billed through Hydrate.

Install See the benchmarks
Five ways to run it no API key · local-first
What changes after install

The same prompt, twice. Once cold, once hydrated.

Left: a fresh Claude Code session, no project context — generic answer. Right: same prompt after Hydrate injects four facts it learned in your previous session (API_BASE, the auth pattern, the data-fetching library, the "do not touch redux" rule). The boundary in the middle is what Hydrate carries across.

Memory transfer between two coding sessions: a new session with no context on the left, the same prompt answered correctly on the right because four extracted facts (API_BASE, auth-pattern, tanstack-query, do-not-touch-redux) were injected from project memory across the session boundary.
Cold vs. hydrated session — facts injected at SessionStart via the Claude Code hook.
Live recording · Sonnet 4.6 · Hydrate v0.4.2

Real Claude Code, no fakery.

Captured live inside the bench/compact-survival-hydrate:v0.4.2 Docker container, the same image that just scored on the compact-survival benchmark. Phase 1 runs a real claude --print against a partial rename (4 of 8 Go files done); the Stop hook captures state to ~/.hydrate/data.db; Phase 2 starts a fresh claude --print with no re-briefing. The model's Phase 2 reply explicitly names the prior task, the work already done, and finishes the remaining files. That's the SessionStart hook injecting the captured state, visible only in the model's response, exactly how a real user experiences it.

Raw asciinema cast available at /demos/recovery-live.cast replay with asciinema play recovery-live.cast for the full session including timing. No editing, no scripting; the model output is verbatim.
Side-by-side terminal comparison. Without Hydrate, 40% of input tokens (4,000 of 10,000) are wasted re-explaining the project (stack, repo layout, feature flags, logging library, test conventions) before the actual request. With Hydrate, the same context is injected concisely from project memory, leaving 100% of the budget for actual work. A bar chart at the bottom shows relative cost: Without Hydrate 100%, With Hydrate 60%, i.e. a 40% lower input token cost with Hydrate.
40% of a typical prompt's input tokens are spent re-explaining the project. Hydrate puts that budget back into work.
Cost calculator

What does Hydrate save your team?

Slide to your team size and plan type. Numbers update instantly. Based on benchmark data - indicative figures for planning.

Full ROI breakdown
20
Save ~£2,900 / month
20 devs · 71% session rescue + cache improvement · indicative
One engine · five surfaces

Five ways to use Hydrate.

Same binary, same memory, same savings. Claude Code, VS Code + Copilot, Codex CLI, Mistral Vibe and MCP all share state. A decision captured in one injects into the next prompt of any other.

Five coding tools (Claude Code, VS Code, Codex CLI, Mistral Vibe, and a generic MCP client showing JSON-RPC) all displaying the same onboarding code, connected to a central glowing memory cube labelled ~/.hydrate/data.db. Blue dashed arrows show bidirectional sync between every tool and the central memory; one orange dashed arrow highlights the currently-active connection from Claude Code.

Claude Code

shipping

Native hooks - UserPromptSubmit, Stop, PreToolUse, PostToolUse. Zero configuration. Works on day one with your existing Claude subscription.

Install guide →

VS Code + Copilot

v1 launch

VS Code extension with @hydrate chat participant plus three Language Model Tools Copilot auto-invokes. 86-95% measured token reduction across ten scenarios.

See the Copilot page →

Codex CLI

alpha

OpenAI's Codex CLI wired into Hydrate via MCP. Shares the same local memory store as Claude Code and Copilot, so you can switch model providers without losing project context.

See the Codex page →

Mistral Vibe

shipping

MCP server wired into Vibe via hydrate vibe install. Receives Claude Code and Copilot project memory on the first tool call. Contributes facts back to the shared store.

Read the test →

MCP server

shipping

Works with every MCP-aware client - Claude Desktop, Cursor, Cline, Zed, Gemini CLI, your own agent. One endpoint, the same memory, the same savings.

Integration guide →
BETA Pricing

Start free. Lock $5/mo Pro for your first 12 months.

Register during beta (whichever edition you pick) and your Pro rate is locked at $5/mo for your first 12 months; $9/mo retail thereafter. How this works →

BETA · locks $5/mo for 12 months

Free

$0/ forever

Up to 2 active projects, full hooks + dashboard. No account. Registering during beta also locks the $5/mo Pro rate for your first 12 months.

Install + lock rate
BETA · $5/mo for 12 months

Pro

$5/ mo · first 12 months

$9/mo retail thereafter · free during beta + 30 days after v1 launch

Unlimited active projects plus per-project backup / restore. Lock $5/mo for your first 12 months of Pro; $9/mo retail after.

See Pro details →
Post-beta

Team

$29/ user / month

Shared project memory across teammates. Indicative launch rate; Team is not part of the beta lock.

Contact sales →

Enterprise

Custom

Self-host or managed. SSO, org-scoped policy facts, audit logs, air-gap support. Online and offline, data synced via git.

Learn more →
Ready when you are

Stop re-explaining your codebase. Start now.

One command. Works in 30 seconds. Local-first, no API key required. 

curl -fsSL gethydrate.dev/install | sh
macOS Linux Windows Works in 30 seconds