Workstations
Local visibility, governance, and optimization for supported AI tools.
Codex
Claude Code
GitHub Copilot
Cursor
Windsurf
OpenCode
The unified control plane for every AI agent
Understand, govern, and optimize AI usage across workstations, production workloads, and hosted agent environments.
Visibility
Understand the business value of AI.
Attribute AI consumption to applications, tasks, and types of work. Connect the cost of a task to what happened to its output, from merged pull requests to discarded code.
Visibility
Know what your teams are using.
See the endpoints, harnesses, models, MCP servers, and skills across your fleet. Explore adoption and usage, and identify assets that need attention.
Budgeting
Put your AI budgets into practice.
Track spend against budgets and forecasts. Define budgets for your teams and choose how TokenShift responds when a limit is reached: warn, downgrade to a default model, or block usage.
Harness Controls
Your AI. Your rules.
Define which harnesses, models, MCP servers, and skills your teams can use. Apply controls in the context of the work, and see their reach and activity.
Optimization
Make AI usage more efficient.
Apply novel token optimization techniques on the endpoint using multiple mechanisms. Give people guidance in their AI workflow to improve how they use agents.
Investigation
Ask a question. Follow the evidence.
Investigate AI spend and usage with Analyst. Ask about teams, models, and types of work, then explore the charts behind the answers and ask follow-up questions.
Wherever AI work happens.
Run TokenShift alongside your AI agents across workstations, production workloads, and hosted cloud environments.
Production workloads
Run alongside production agents in cloud containers or through agent SDK integrations.
- Cloud containers
- Agent SDKs
Hosted agents
Install in your agent's cloud runtime through its environment setup.
- Claude Code Cloud
- Codex Cloud
Closer to the work. More control.
Native harness hooks and a local gateway put TokenShift where context is created and decisions are made. Optimize locally, with visibility and policies managed centrally.
Private by design.
Prompts, responses, and file contents are never sent to PointFive's control plane. Analysis runs locally; only derived metadata, such as token counts, timings, and attribution, reaches the console. There is no option to collect prompt content.
Optimize at the endpoint.
TokenShift combines novel token optimization techniques across multiple mechanisms, running directly on the endpoint to make AI usage more efficient.
See the work behind the tokens.
Classify token usage as personal or work, identify the type of work, and attribute it to an application or project. Expertise-level scoring adds a deeper understanding of how your teams use AI.
Make policy part of the workflow.
Set policies centrally and explain them where developers work. In-session guidance shows why a policy applied and points to the compliant next step.
Scale with your team.
Local processing uses the machines you already have, with no shared inference gateway to size or operate. One package detects supported harnesses and configures their hooks and local gateway.
Keep work moving.
If local optimization errors or exceeds its time budget, the harness continues with the original input. Processing is isolated per endpoint, so one machine's issue does not interrupt the whole team's AI work.