Skip to main content
RUSTGRID
Run one issue free

Self-hosted coding agent

Run coding agents on your infrastructure with visible controls.

Keep execution near your code and tools. RustGrid coordinates the issue, worker lease, run history, checks, and pull request handoff without hiding the operational path.

Run one issue free →

Your worker. Your execution environment. A visible review path.

One complete self-hosted run

Issue #631 becomes pull request #638.

A concrete, fictional example showing an issue dispatched to a connected worker, validated on the same infrastructure, and handed back for review.

acme/platform · worker eu-build-02PR ready
security

Redact secrets from worker diagnostics

Diagnostic bundles must mask credential-like environment values while retaining the keys operators need for support.

Acceptance
Known secret patterns are redacted in text and JSON output.
Execution
Linux worker in the team’s private build network.
01
Worker claimed the runLease issued to eu-build-02
✓
02
Agent changed 4 filesRedactor, serializers, fixtures, tests
✓
03
Local gates passed41 tests · audit fixture scan · lint
✓
04
Pull request #638 opened+143 −19 · security review requested
→
Controlled resultThe code ran on the connected worker; the evidence returned with the PR.

Run events show where execution happened, what was validated, and which human review is still required.

✓Redaction tests41 passed
✓Fixture scanclean
·Security reviewwaiting

Visible boundaries

Control where code runs and see what the run did.

Self-hosting gives you the execution boundary. RustGrid supplies the durable coordination around it: assignments, leases, events, validation status, and review state.

Your infrastructureExecution boundary
  • Worker host and operating system
  • Repository checkout and build tools
  • Network access and internal dependencies
  • Workspace retention and cleanup policy
RustGrid control planeWorkflow boundary
  • Issue context and execution manifest
  • Worker assignment and run lease
  • Observable events and quality gates
  • Pull request and human review state

Operational controls

A worker is part of the system, not a hidden subprocess.

Explicit assignment

A compatible worker claims a server-owned execution with a bounded lease.

Visible
Run telemetry

Lifecycle events, validation, failures, and review state remain attached to the issue.

Visible
Operator policy

You configure the worker host, toolchain, credentials, network, and cleanup posture.

You control
Human decision

The resulting pull request waits for an authorized reviewer rather than merging itself.

Required

The honest boundary

Self-hosted does not mean risk-free.

You remain responsible for hardening the host, limiting credentials and network access, protecting retained workspaces, and choosing an isolation model that matches your code and threat model. RustGrid makes the execution path observable; it does not erase operator responsibility.

Review the security boundary →

Prove the path with one issue

Connect your worker. Keep the run visible.

Start with one scoped task and review the resulting pull request before expanding the workflow.

Run one issue free →