Private evaluation · Context search engine

Context your agents can use

Index the codebase. Remember what the team learned. Search both in one question. Models are optional.

Get started
A query in a configured evaluation
shepaird call context.search --input '{
  "query": "Where is refund eligibility checked?",
  "org_id": "YOUR_ORG_ID",
  "actor_id": "YOUR_ACTOR_ID",
  "budget_tokens": 2000,
  "limit": 10,
  "explain": true
}'

An example query after your evaluation environment is configured.

How it works

Index your code. Retrieve context. Save findings. Brief the next session.

01 Index

Map the codebase

Build a searchable index of your repositories and refresh it as the code changes.

02 Search

Ask one question

Search code and saved knowledge together, with available source references.

03 Remember

Save what you learned

Attach citations so findings can be checked as the code evolves.

04 Brief

Start the next session ready

Open with in-progress work, the memories that matter, and the last handoff.

Code, memory, and search. One product.

Code

A live map of the codebase, so agents do not re-read thousands of files every session.

Memory

Saved team knowledge with provenance, citations, and verification status.

Search

One question. Code and memory come back together, ranked, with citations.