More intelligence per token.
Glean’s agent harness improves quality and reliability while reducing waste with better context management and efficient execution so agents can take on harder work without adding token waste.
Model choice, not lock-in
Glean's agent harness supports frontier and open-source models, so you always run the most efficient one for the job.
Built for token efficiency
Glean's harness calls tools more efficiently through code, uses only the context each step needs, and reserves larger models for work that requires them.
Optimized for enterprise work
Glean’s agent harness is built for enterprise work, from data analysis to artifact generation to automations, with thousands of evals run each quarter to hold the quality bar.
Glean Waldo
Frontier intelligence, without the frontier cost.
Not every task needs a frontier model. Waldo, Glean’s agentic search model, reduces latency by 50% and tokens by 25%, reserving advanced models for work that needs them. Glean’s agent harness routes work across 30+ models, including Claude, Gemini, OpenaAI, and NVIDIA, with no provider lock-in.
Coding harness
Execute in code, stay efficient at scale.
Glean’s agent harness uses code as its execution layer. Tools are exposed as callable functions, logic runs in a sandbox, and intermediate data stays in variables and files instead of re-entering the model at every step. The result is 24% fewer tokens, faster execution on complex multi-step work, and a model-agnostic layer that can swap models freely.
Context management
Context that keeps AI on track.
Glean’s agent harness keeps the context for the task narrow, selecting, trimming, and structuring only what each step actually needs. Agents stay focused and coherent across multi-step workflows, without crowding the model’s context window with unnecessary noise.
Trace learning
AI that learns how work gets done.
Trace learning captures what worked and what didn’t during execution, then turns successful tool sequences and workflow patterns into reusable strategies. Over time, the harness automatically improves for your team’s workflows without manual configuration.
Explore more resources
The token economy: How enterprise AI architecture impacts cost and utility at scale
The token economy: How enterprise AI architecture impacts cost and utility at scale
Glean Waldo: An agentic search model that pairs with LLMs to deliver frontier intelligence with ~50% lower latency and ~25% fewer tokens
Glean Waldo: An agentic search model that pairs with LLMs to deliver frontier intelligence with ~50% lower latency and ~25% fewer tokens
Building an efficient harness for advanced enterprise work
Building an efficient harness for advanced enterprise work
The token economy: How enterprise AI architecture impacts cost and utility at scale
The token economy: How enterprise AI architecture impacts cost and utility at scale
Glean Waldo: An agentic search model that pairs with LLMs to deliver frontier intelligence with ~50% lower latency and ~25% fewer tokens
Glean Waldo: An agentic search model that pairs with LLMs to deliver frontier intelligence with ~50% lower latency and ~25% fewer tokens
Work AI that works.
