Agent8088 / Palindrome Research Labs
Reliable AI agents—even on small,local models.
An open-source agentic harness that prioritizes small local models for reliable, verifiable task completion.
Agent8088 / launch demo
hover for sound
Agentic harness
Agent8088
The small-local-model-first open-source agentic harness for reliable and verified task completion.
Resolves real pain points through research-backed approaches.
Transparent, sandboxed, safe, in your control. Execution grounded. Robust and context aware.
Download
56 built-in tools12 provider profilesMCP both waysOS-level sandbox6 sub-agent profilesLocal hybrid memory8 surfaces, one gateProvider fallbackCron-scheduled runs10 security layersSSRF-guarded egressPlan-audit verificationGit push always blockedDaily JSONL audit logTurn-budget bounded
The installer provisions an isolated Python environment and installs the global agent8088 command. No administrator access required for the base install.
Why Agent8088
Why Agent8088finishes the task.
Resolves real pain points through research-backed approaches. Transparent, sandboxed, safe, in your control. Execution grounded. Robust and context aware.
Verified completion
- Reliable task completion using small local models
- Verification-gated task completion checks
- Execution-grounded verification
- Reliable control loop
Context aware
- Context awareness, pre-call budget and overflow prevention
- Robust context management
- Tool minimization and optimization
Sandboxed and safe
- Permissions and security handling
- OS-level sandbox
- Usage monitoring and transparency
Robust, low friction
- Failure recovery
- Focused on solving common user pain points using research-backed approaches
- Feature rich
- Low friction
- Extensible