A prompt-injected model cannot skip your approval step
How 10xGraph keeps handoffs, tool visibility, tool results and interrupt resumes under the graph's control, so a steered model cannot walk around a human approval.
Blog / category
How 10xGraph works inside and why: failure modes, storage, auth and the runtime.
How 10xGraph keeps handoffs, tool visibility, tool results and interrupt resumes under the graph's control, so a steered model cannot walk around a human approval.
When an AI run belongs in a request and when it belongs in a job, and how 10xGraph runs jobs as a thin TaskIQ layer with a worker, a scheduler and the same auth.
The limits a 10xGraph run has by default: parallel tool cap, run, node and tool timeouts, per-user run cap, one run per thread, request size limits and paging caps.
What 10xGraph keeps out of logs, traces and Sentry: hashed user ids, redacted Redis URLs, tool error references, escaped client values and generic errors in production.
What auditors ask of an AI backend, which controls 10xGraph supports in software, and which stay with your organization. Includes a short team checklist.
How fetch_url and client-supplied media links are limited to public addresses in 10xGraph: pinned connections, checked redirects, MediaFetchPolicy and a pre-run 400.
Follow one request through 10xGraph: API server, access control, data and state, background jobs, observability and deploy. What you write and what runs for you.
What 10xGraph runs around your agent by default, the config that turns each protection on, what 1.0 adds, and the honest list of what stays your job.
How 10xGraph keeps users and tenants apart for teams shipping AI to their customers: verified identity, roles with reach, 404 outside reach and tenant columns.
Gate pull requests on what an AI agent did: which tools it called and with what arguments. Uses 10xGraph eval sets, trajectory scoring with no judge model, and pytest.
Cheaper models, a stateless MCP spec, MCP CVEs, approvals everywhere and EU transparency rules: what changed in AI in 2026 and the 10xGraph code to handle each one.
Why a crash mid-node re-runs tools that already fired, why the usual fixes fall short, and how 10xGraph's tool ledger and versioned writes close the gap.
How PgCheckpointer splits agent state between a Redis hot cache and versioned Postgres history, keeps them consistent, and behaves when either one fails.