One agent run helps every teammate
When a coding-agent session ends, the useful context is chunked, embedded, and shared so the next engineer can build from it instead of rediscovering it.
Shared memory for agentic engineering
Share your agents' knowledge fromClaude CodeCodexOpenCodeCursor agentsDevin
Honeygraph captures what coding agents already learned across Claude Code, Codex, OpenCode, Slack, API specs, URLs, and notes, then shares it with your teammates before their next answer.
12.8k
shared chunks stored
5
agent hooks installed
1.5s
team memory lookup
Claude sessionsCodex runsOpenCode workSlack messagesOpenAPI specsURLsplain textsolutionscommandsproject contextteammate fixesshared memorysemantic searchClaude sessionsCodex runsOpenCode workSlack messagesOpenAPI specsURLsplain textsolutionscommandsproject contextteammate fixesshared memorysemantic search
Built for engineering teams
Every expensive agent run should teach the whole team.
Team memory returns before the next answer
Prompt hooks query shared memory and inject relevant teammate context back into Claude Code, Codex, OpenCode, or whatever agent interface your team uses.
More than one person’s chat history
Slack messages, OpenAPI specs, URLs, plain text, and structured solutions all enter the same team index for semantic search.
The Honeygraph flow
From spent tokens to shared team memory.
01 / capture
Work drops pollen automatically
The CLI hooks into agent workflows to collect sessions, prompts, tool activity, commands, and final transcripts without changing how engineers work.
Claude CodeCodexOpenCodeCLI hooks
02 / embed
Knowledge is chunked and indexed
The cloud API parses sessions, Slack messages, OpenAPI specs, URLs, and text, then stores embeddings in Aurora with pgvector.
sessionslackopenapihttppgvector
03 / inject
Teammate knowledge appears in the next prompt
Queries run at session start and prompt submit. High-scoring team knowledge is injected as context before the agent responds.
queryscoreteammate fixcontext
honeygraph hook installready
$ curl -fsSL https://honeygraph.dev/install | sh
$ honeygraph auth login
$ honeygraph hook install --agents claude,codex,opencode
Installs agent hooks so every supported coding assistant can contribute to and retrieve from the same shared team memory.
Local CLI, shared cloud memory
Install once. Your teammates start benefiting from the work already happening.
The CLI handles auth, agent hooks, manual ingestion, querying, skills, doctor checks, and upgrades. The cloud API handles parsing, embeddings, storage, and team-wide semantic search.
Shared operational memory
Ask what your teammates already solved.
claude-codefound teammate context for this repo
honeygraph run SessionStart3 team hits injectedcodexcaptured commands and review context
honeygraph ingest --current-sessionqueued for embeddingslacksaved deployment note from teammate
Save to Honeygraphembedded as team knowledgeopencodereused a prior fix from another teammate
honeygraph query retry backoffteam memory returnedStop making teammates pay twice for the same agent lesson.
Honeygraph turns agent sessions, teammate fixes, Slack context, docs, and specs into shared memory so every future prompt can start with what the team already learned.
hello@honeygraph.dev