Rift Code / Local AI memory

The code changed.Keep the story.

A local record of AI coding sessions, tied to the repository. See what happened, verify the history, and give an agent the context that changed.

Explore the record
One captured sessionIllustrative
01PROMPT
02EDIT
03CHECKPOINT
Agent actionChanged hunksVerifiable history

SAMPLE RECORD / 09:41–09:46

Claude Code3 hunks / 2 filesCheckpoint verified

01 / Capture

A record that travels with the repository.

Rift records supported agent sessions, edits, usage, and checkpoints locally. Sensitive material is redacted before durable storage. The append-only Git-native ledger makes later changes detectable.

SESSIONClaude Code
CHANGEChanged hunks
RECORDCheckpoint

Illustrative sequence, not a live session

Example / Captured turn

One AI turn, captured end to end.

Follow an illustrative request from prompt to inspectable record. Select a stage to see what Rift keeps and where Recall can help.

Session 01 / Illustrative

Developer

A change is requested.

The developer asks a supported coding agent to add a rate limit to an API route.

Captured view / Prompt

09:41 / CLAUDE CODE
› Add a token-bucket limit to the gateway route
repository / api-gateway

The prompt and subsequent tool activity become part of the captured session.

02 / Recall

Remember what the agent already saw.

Recall checks a known state against the current repository and returns a bounded answer. Unchanged context can stay out of a repeat read; changed context can be brought forward.

KNOWNPreviously read
CHECKStill valid?
DELIVEROnly the needed change

Savings shown in Rift are estimated upper bounds, not verified billing reductions.

Scenario / Recall

What could repeated context cost your team?

Explore a possible Recall reduction. This is a projection, not measured savings or a guarantee.

Your team

25

People making AI coding requests on a typical workday.

Turns per developer / day
$0.2400

Illustrative turn: 50k input + 2k output tokens at the selected model’s standard API price.

Projected Rift savings / month

$1,852

$22,226 annualized at this rate · 49% of modeled spend

Modeled monthly spend$3,780
With assumed Recall reduction$1,928

Assumes 21 workdays, 70% of each turn’s spend tied to repeated context, and Recall avoids 70% of that amount. Caching and discounts are excluded; Rift does not verify bill savings. Standard API list rates checked 30 September 2026; Gemini Flash introductory rates run through December 2026. Agent subscriptions and tool charges may differ. View selected model pricing.

03 / Inspect

Read the work behind the code.

Canopy, Rift's terminal dashboard, brings sessions, checkpoints, recall, and cost into one place. Claude Code, Codex, and Antigravity are the currently wired capture integrations; coverage varies by agent.

CANOPY / REPOSITORY
sessions → checkpoints
recall → context state
usage → local view

Rift Code

Give the next session a better starting point.