Context infrastructure
Context infrastructure
for AI agents.
Starting with high-stakes legal workflows, where context can change the decision.
Built with support from
for Startups
Activate
AltaLab
AI agents are getting better at reasoning.But better reasoning doesn't fix incomplete context.
An agent can make the wrong decision even when its reasoning is sound, because it never had the information the decision depended on.
What the agent saw
What the decision needed
- MSA
- Amendment 1
- Amendment 2Never seen
- Amendment 3
- Schedule
- Counsel email
Know if your agent has enough context to execute.
Tensor is GammaLex's context runtime. One call returns the evidence and tells your agent whether it has enough context to make a better decision before it executes.
The agent sends a request
One call, before it executes: who is asking, and about what.
GammaLexAgentSDKTensor works out what it needs
What must be known, as you define it, and what the agent may use.
GammaLexOrchid MSA — Project LanternBefore any readRequirements
- Executed agreementRequired
- Amendment chainRequired
- Change-of-control clauseRequired
- Disclosure scheduleRequired
- Counsel adviceRequired
412 records permitted · 1,906 excluded
It gathers the evidence
Only what the request depends on, from the sources the agent may read.
GammaLexcontext6 selectedSelected context
- Executed MSA, Calder ↔ Orchid Cloud LtdData room
- Amendment No. 3 (2026-03-18)Contract repository
- Disclosure schedule item 7.3 (2026-02-03)Disclosure schedule
- MSA §14.2 — change of controlData room
- Outside counsel advicewithheld
4 of 6 shown · 1 withheld
It connects the records
How the records relate, where they conflict, and what is missing.
GammaLexcontext graph8 recordsHow the evidence connects
MSA §14.2 — change of controlConflictData room
Requires counterparty consent on a change of control.
It shows whether the agent has enough context to execute
and what stands in the way when it doesn’t.
GammaLexRUN_7D41C2CheckedRequirements
- Executed agreementEstablished
- Amendment chainNot found
- Change-of-control clauseContradictory
- Disclosure scheduleStale
- Counsel adviceRestricted
1 of 5 established
It learns from every run
Each run is remembered, so the next request starts from what's already established.
GammaLexOrchid MSA — Project Lantern3 runsEach run builds on the last
Amendment No. 2 was added to the data room.
- Executed agreementCarried forward
- Amendment chainNewly established
- Change-of-control clauseStill open
- Disclosure scheduleStill open
- Counsel adviceStill open
2 of 5 established
One complete run.
The same run, as the agent receives it.
GammaLex AIContext / Runs / RUN_7D41C2Can we rely on the Orchid MSA change-of-control position for the diligence memo?
Diligence agent · Orchid MSA — Project Lantern · via SDK · 6 records
To resolve
- Provide Amendment No. 2 (Data room)
- Confirm which governs: MSA §14.2 or Disclosure schedule 7.3
- Counsel advice is privileged — request access or a summary from the matter lead
What Tensor returns.
Context Manifest
The context assembled for this request, structured for your agent to use immediately.
Evidence
The information the request depends on, gathered from permitted sources.
Context sufficiency
Whether the agent has enough context to execute, and why.
Boundaries
What Tensor could and could not use, with boundaries preserved.
Trace
What Tensor used and how the context was assembled.
Not an LLM. Not RAG. Not guardrails. The context layer in between.
Before your agent executes, GammaLex determines what it needs to know and whether it has enough to make better decisions. It works with any model or framework.
LLM
Reasons
Thinks with what it is given.
RAG
Retrieves
Finds what looks relevant.
Guardrails
Constrains
Limits what AI can say or do.
GammaLex
Contextualizes
Determines what the agent needs to know and whether it has enough to make better decisions.