BackContext

One context layer that records every agent’s work and outcome. BackContext keeps an append-only journal of every session, your skills, your keys and your approvals in one place, and turns verified outcomes into training and eval data for your next model. Read more at backcontext.com.

Engagement

What’s included in this service?

Journal

An append-only, attributed record of every session from every tool: who did the work, on which device, and what came of it.

Versioned skills

Store each SKILL.md once and sync it into Claude Code, Codex, Cursor, Grok and Gemini CLI. Every version is kept.

Vault and approvals

Keys encrypted at rest with AES-256-GCM and released per policy. Each request shows tool, device and purpose; approve or deny it from the dashboard.

Audit trail and CLI

Every request, approval and release is written down. One binary, --json everywhere, MCP server built in; init wires up installed tools in about 2 minutes.

Why it matters

Why this service matters

Every model we train starts as a record in BackContext. Customers have captured 665K+ work records so far, and weekly record growth went from 24K to 210K between August and September 2026. Each verified outcome becomes training and eval data for the next model version.

Running across agent tools in a day

665K+ work records captured

Verified outcomes become training data

Every request on the audit trail

FAQ

Frequently asked questions

Have more questions?

Book a call