Molten
Governed agent runtime: kill switch, circuit breakers, four-tier sandbox, dead-man's switch. Private build with no public artifacts.
Project details
- Classification
- Private build
- Repository
- Not public
- Language
- TypeScript
- License
- Private
- Status
- Paused
- Tests
- 1 test notes
- Last verified
- 2026-08-06
This is my private build. I describe its design below, but there are no public source files or test results you can use to verify the implementation.
What it is
Molten is my private agent runtime with stop controls and execution limits. I describe its design here, but there is no public repository or test output you can use to verify the implementation.
Architecture
- Kill switch terminates a run from outside the agent's own control loop.
- Circuit breakers trip on repeated tool failure instead of retrying indefinitely.
- Four-tier sandbox escalates capability only as the run earns it.
- Dead-man's switch halts execution when the operator heartbeat stops.
Tests and gates
- No public test artifacts. Test claims for this build are not independently verifiable.
Limits
- Private build. No repository URL, no package, no public CI.
- Described here for architecture context only.
Related methods
Acceptance tests for agent work →
Executable checks test the task and detect regressions.
Evaluation tests for releases →
A fixed evaluation suite checks release requirements.
MCP tool permissions →
The server checks which tools a caller may use.
Recording agent runs →
A shared trace ID connects the records of a run.
Spending limits for agent runs →
Budget checks run before provider requests.
Pausing for human approval →
The run waits for approval before the action executes.
Related projects
ClawKeeper →
ClawKeeper checks proposed Claude agent actions against explicit rules before a model call. Autonomous spend defaults to zero, and the database enforces an append-only audit log.
TechTide Harness Kit →
TechTide Harness Kit uses JSON Schema to declare skill permissions, risk tiers, network access, and approval requirements. Automated checks validate declarations and run prompt-injection fixtures.
CipherClaw →
CipherClaw analyzes recorded agent activity without model calls. It builds causal graphs, scores possible causes, and flags failure patterns.
Swarm 357 →
357 Claude agent roles across six business layers, with portable memory, enforced cost caps, a bash policy gate, and durable checkpoints.
I document the project here. TechTide AI publishes current service scope, prices, and terms.
Systems Decision Audit at TechTide AI→© 2026 Alex Cinovoj · TechTide AI · Columbus, OH