# julep-ai/julep

Julep — durable, composable AI agents. Flows that crash and resume, retry safely, and explain every step.

Repository: https://github.com/julep-ai/julep
Canonical: https://ross.abutalabs.com/products/julep
Homepage: https://docs.julep.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agents, ai-agents, ai-agents-framework, ai-memory, ai-platform, ai, developer-tools, llm, llm-ops, python, aiagents, devfest, node, node-js, nodejs
Last push: 2026-08-06T05:17:30+00:00

## Health v2 (maintenance only)
Score: 68/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 35, longevity 62
- inputs: {"age_days": 881, "days_push": 27, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6591, forks 970 (observed 2026-08-28T04:09:45.536402+00:00)

## What it is
Julep is an open-source framework and platform for building durable, composable AI agents as dataflows that can crash, resume, retry, and explain every step. It includes a Python SDK, a Node.js SDK, a Temporal-backed workflow engine, long-term memory, MCP integration, and a self-hosted OpenAI Responses API-compatible server.

## Use cases
- build ai agents that remember past interactions
- run long-running multi-step llm workflows that survive crashes
- orchestrate parallel branches and tool calls in an agent pipeline
- self-host an openai responses api compatible server
- connect agents to mcp servers and external tools
- add retries and timeouts to llm tool calls
- create stateful chatbot sessions with persistent memory

## When to choose
- you need durable, resumable agent workflows rather than simple prompt chains
- you want long-term memory and stateful sessions across interactions
- you need strict control over which tools an llm may call
- you want a self-hosted, open-source alternative to managed agent platforms

## When to avoid
- you only need a single prompt-to-response llm call with no state
- you cannot operate the Temporal/PostgreSQL infrastructure the platform relies on
- you need a fully stable 1.0-style API and cannot tolerate release-candidate changes

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, workflow-automation, llm-inference, mcp, webhook, scheduling
- domain: large-language-models, developer-tools, self-hosted
- platform: python, self-hosted, cross-platform
- tags: durable-execution, temporal, ai-workflows, long-term-memory, tool-calling, open-responses-api, multi-step-tasks, ai-agents, automation, nodejs, docker

## Member repositories
- julep-ai/julep (main) score 68

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:45.536402+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:43:33.833097+00:00, confidence not recorded.
  - readme: https://github.com/julep-ai/julep (fetched 2026-08-28T04:09:45.536402+00:00, sha 00f45d8b3e25)
  - homepage: https://docs.julep.ai (fetched 2026-08-29T08:40:07.021260+00:00, sha 9a0d1e350285)
  - site_page: https://docs.julep.ai/responses/quickstart (fetched 2026-08-29T08:40:07.023845+00:00, sha 200944e6cc85)
  - site_page: https://docs.julep.ai/integrations/supported-integrations (fetched 2026-08-29T08:40:07.025795+00:00, sha 9b7aadfad5b7)
  - site_page: https://docs.julep.ai/introduction/install (fetched 2026-08-29T08:40:07.027345+00:00, sha ecbd1c75d292)
  - site_page: https://docs.julep.ai/introduction/quickstart (fetched 2026-08-29T08:40:07.028896+00:00, sha 2eb20a37787e)
  - site_page: https://docs.julep.ai/FAQ (fetched 2026-08-29T08:40:07.030399+00:00, sha c06a3aaedfae)
  - site_page: https://updates.julep.ai (fetched 2026-08-29T08:40:07.033037+00:00, sha a3280ec24b68)
- Data as of 2026-08-30T08:39:29.467469+00:00.