mem0-integrate

mem0-integrate

Wire Mem0 into an existing repo with a goal-driven, test-first pipeline. Pairs with mem0-test-integration for verification.

Canonical sources (fetch before deciding anything)

The skill MUST WebFetch these URLs before step 3 and cite them in plan.md. They are the ground truth — do not rely on ambient knowledge of the Mem0 API.

Agent-ready docs

• Scope-tagged docs index: https://docs.mem0.ai/llms.txt
• Full docs (single file, deep dives): https://docs.mem0.ai/llms-full.txt
• OpenAPI spec (Platform REST, machine-readable): https://docs.mem0.ai/openapi.json
• Hosted MCP server: https://mcp.mem0.ai (requires Platform API key)
• Integrations index: https://docs.mem0.ai/integrations

Published Mem0 skills — delegate; do not reimplement

Prefer these over writing your own call-site patterns. Each is a standalone SKILL.md with triggers, examples, and version-pinned code.

• SDK (Python + TS, Platform + OSS): https://raw.githubusercontent.com/mem0ai/mem0/main/skills/mem0/SKILL.md
• CLI: https://raw.githubusercontent.com/mem0ai/mem0/main/skills/mem0-cli/SKILL.md
• Vercel AI SDK: https://raw.githubusercontent.com/mem0ai/mem0/main/skills/mem0-vercel-ai-sdk/SKILL.md
• Editor/MCP plugin glue (9 MCP tools): https://github.com/mem0ai/mem0/tree/main/integrations/mem0-plugin

SDK source (read when docs are ambiguous)

Public repo. Cross-check against the mem0_tested_versions range in this skill's frontmatter if the main branch has moved past a major.

• Repo root: https://github.com/mem0ai/mem0
• Python SDK: https://github.com/mem0ai/mem0/tree/main/mem0
• TypeScript SDK: https://github.com/mem0ai/mem0/tree/main/mem0-ts

Quickstarts (for bootstrapping unfamiliar stacks)

• Platform: https://docs.mem0.ai/platform/quickstart
• OSS Python: https://docs.mem0.ai/open-source/python-quickstart
• OSS Node: https://docs.mem0.ai/open-source/node-quickstart
• Platform vs OSS comparison: https://docs.mem0.ai/platform/platform-vs-oss

Integration principles (non-negotiable)

The true goal of this skill is to produce a PR the maintainers can accept without argument. That rules out anything invasive.

1. Additive, not replacing. If the target repo already has a memory system, a session store, a user-context layer, or anything named Memory / memory_*, Mem0 sits alongside it, not in place of it. The existing system keeps working unchanged.
2. Opt-in by default. Gate all new Mem0 code behind a feature flag (env var like MEM0_ENABLED=1, a config key, or a strategy selector). With the flag unset, behavior is the repo's original behavior, byte-for-byte.
3. No breakage. No removed exports, no renamed public functions, no changed method signatures, no modified existing tests, no changed behavior of existing tests. All pre-existing tests must pass unchanged both with the flag set and unset.
4. Minimal dependency surface. Add mem0ai (plus any deps the delegated skill requires) and nothing else. No new vector stores, no graph databases, no provider SDKs the repo does not already use.
5. Separable commits. Code, tests, and config/docs land in separate commits so reviewers can cherry-pick.
6. The null hypothesis wins. If no additive, gated fit exists after step 6 (plan), exit with code 1 and a rationale. A bad PR is worse than no PR.
7. Backend only. Mem0 integration lives in server-side code. API keys, memory scope, and user-identity resolution are not safe client-side. If the repo has both backend and frontend, the call sites live in backend files. Frontend-only repos are rejected at preconditions.

Enforced at four gates: preconditions (reject frontend-only repos and repos where additive fit is impossible), step 2 comprehension (confirm a backend exists and name candidate surfaces), step 6 plan review (reject plans that mutate existing exports or name client-side call sites), and step 10 self-healing loop (refuse to "fix" principle violations — surface them instead).

Skill delegation rules

Before writing any code, check whether a published skill already covers the target stack. If yes, delegate — copy its call-site pattern into plan.md and into the tests; do not paraphrase.

Record the delegated skill's raw URL in plan.md under a "Delegated skill:" field. The test writer in step 7 and the implementation subagent in step 8 both read this field.

Preconditions

Refuse to start unless ALL of the following are true:

• Current working directory is inside a git repository with a clean index (no uncommitted changes). Protects the user's work — every edit lands on a feature branch, not on top of in-progress changes.
• Repo has a detectable language (package.json / pyproject.toml / requirements.txt). No language → exit cleanly with a written rationale.
• Repo has a backend. Detected by: a backend/ or server/ or api/ directory; a Python package with FastAPI/Flask/Django/Starlette; a Node package with Express/Fastify/Koa/NestJS/Next-API-routes; an agent-loop framework (LangGraph, LangChain, LlamaIndex, Agno). Frontend-only repos (pure React/Vue/Svelte SPAs, static sites, mobile-only) → exit with code 1 and a rationale. Mem0 is not installed client-side.
• The user has already decided Mem0 fits this repo. This skill does NOT survey the codebase to justify fit — bring a concrete goal. (Step 2 does read the repo to understand what it does and locate backend integration surfaces; that is mechanics, not fit-justification.)

Exit with a written rationale if any precondition fails. Do not try to "make it work anyway."

Pipeline

1. Language detection

Monorepo with both → ask which subdirectory to operate in, then recurse.

2. Repo comprehension — what does this repo do, and where is the backend?

Before any decision (product, goal, plan), understand the repo enough to locate where in the backend the integration belongs. This is not fit-surveying — the user already decided Mem0 fits. This is mechanics: you cannot write a plan without knowing what files matter.

Read, in order, with a token budget — do not scan the whole tree:

1. README.md (root) + first-page of any README_*.md variants.
2. CONTRIBUTING.md / AGENTS.md / CLAUDE.md at root if present — these often spell out architecture and entry points.
3. package.json / pyproject.toml scripts + entry points.
4. The layout of the top two directory levels (not recursive).
5. Key config files: docker-compose.yml, Dockerfile, Makefile, langgraph.json, next.config.*, nuxt.config.*.

Produce .mem0-integration/repo-summary.md:

# Repo comprehension

**What this repo does:** <one paragraph in plain English. Who is
the end user? What does the app do for them? What LLM / agent
behavior is central? Do not list dependencies — describe behavior.>

**Architecture at a glance:**
  - Backend: <path(s), framework, primary entry point>
  - Frontend: <path(s) if any, framework — for context only; no
    integration here>
  - Agent loop / orchestration: <LangGraph? custom? none?>
  - Existing memory/session/state systems: <name them — these are
    what step 6 Coexistence must preserve>

**Candidate backend integration surfaces** (ranked, best first):
  1. `<backend-file>:<line_range>` — <function> — <one-sentence
     reason this is where write/read could slot in without
     replacing anything existing>
  2. ...
  3. ...

**Not a fit here:** <list anything the skill considered but ruled
out — e.g., "frontend chat component: client-side, excluded by
backend-only rule"; "existing memory subsystem X: would require
replacement, excluded by additive principle">

**Sources read:** <list the files actually opened, with line counts,
so reviewers can verify coverage.>

Show the user the rendered summary and ask: "Is this understanding correct? Which of the candidate surfaces (1, 2, 3 ...) should step 3 forward target?"

Gate rules:

• If no backend surface is found → exit code 1. The preconditions should already have caught frontend-only repos; reaching this point means a more subtle miss (e.g., the "backend" is actually just a static build). Do not force a fit.
• If every candidate surface would require replacing an existing memory/session system → exit code 1 with the "additive principle" rationale. The user can manually point at a non-conflicting location and re-run.
• User corrections update repo-summary.md and re-confirm. Max 3 rounds; beyond that, exit code 1.

The user's chosen surface index is baked into product.json as preferred_site and referenced by steps 5 and 6.

3. Product selection — Platform vs OSS (ask with a recommendation)

Read the ## Identify the User's Setup block in https://docs.mem0.ai/llms.txt for the Platform-first routing rules, then apply the heuristics below. Ask, but never blank:

• Other managed-service SDKs present (@clerk/*, stripe, @supabase/*, openai, @upstash/*, posthog-*) — 3+ → recommend Platform.
• Local-infra signals (docker-compose.yml with postgres / redis / qdrant / neo4j, ollama configs, self-hosted auth) — 2+ → recommend OSS.
• No strong signal → default recommendation: Platform (lower integration cost; migration later is supported).

Example:

I see stripe, @clerk/nextjs, and @supabase/supabase-js — managed services throughout. I recommend Mem0 Platform (4-line integration). Override and use open source?

Bake the choice into the goal doc in step 5. Do not re-decide later.

4. API key check (env-first, then ask)

If present in env → continue. If MEM0_API_KEY is missing AND the track is Platform → default to Agent Mode: run mem0 init --agent --agent-caller <your-name> --json (after pip install mem0-cli or npm install -g @mem0/cli), substituting your agent identity (e.g. claude-code, cursor, codex). If you forgot to pass --agent-caller, run mem0 identify <your-name> after init. Cache the key to .env (with user consent) and continue. Tell the user to claim later with mem0 init --email <their-email> — same key, no agent disruption. If missing AND CI mode (MEM0_INTEGRATE_CI=1) → exit with code 2 and the name of the missing key.

Never echo key values into trace.jsonl. Persist to .env only with explicit user consent, and append .env to .gitignore if not already there.

If the user is on OSS and wants a non-OpenAI LLM, route them to the components/llms/* docs and re-run this step with the chosen provider's key.

5. Goal doc — the hard gate

Write .mem0-integration/goal.md and require user approval before step 6.

Template:

# Mem0 Integration Goal

**What gets stored:** <one sentence — user utterances? extracted
preferences? a specific domain fact like "dietary restrictions"?>

**When it gets retrieved:** <one sentence — on each user turn? before a
specific tool call? at session start?>

**Why:** <one sentence — the user-visible behavior change. "Assistant
remembers previous orders across sessions," not "we added memory.">

**Product:** Platform | OSS  (locked from step 3, do not change)

**Delegated skill:** <raw URL of the published skill being used
from "Skill delegation rules" above, or "none — custom integration
against `skills/mem0`">.

**Out of scope:** <anything explicitly excluded: "no graph memory,"
"no multimodal," "no migration from existing store">

Rules:

• User must approve explicitly. If they edit the doc, reload and re-confirm.
• goal.md is the contract the test suite is written against. Never rewrite it after step 6 starts.
• Max 3 rejection rounds. On the 4th, exit with code 3 and the rejection notes — the integration is not well-specified enough to proceed.

6. Integration plan — how and where (hard gate)

Given goal.md is "what and why," this step produces "where and how" and gets explicit user sign-off before any code is written.

The skill does a scoped read of the repo (no wide survey):

• Grep for the LLM call sites that match the goal (e.g., openai.chat., anthropic.messages., model.generateContent, ChatOpenAI, createLLM).
• Grep for the user-identity source (req.user, session.user, auth(), ctx.userId, cookies).
• Check package.json / pyproject.toml / requirements.txt for conflicts (e.g., existing mem0ai at a different version).

Then write .mem0-integration/plan.md:

# Mem0 Integration Plan

**Write pattern:** <one sentence — e.g., "After each assistant reply,
call client.add([user_msg, assistant_msg], user_id=<source>).">

**Read pattern:** <one sentence — e.g., "Before building the LLM prompt,
call client.search(query=latest_user_msg, user_id=<source>, limit=5)
and inject results as a system message.">

**User identifier source:** <code path — e.g., `req.auth.userId`,
`session.user.email`, `ctx.params.user_id`. If none, ask the user.>

**Session scoping:**
  - user_id: <source>
  - agent_id: <static slug | null>
  - run_id:   <source | null>

**Write call site:** `<file:line_range>` — inside `<function>`
**Read call site:**  `<file:line_range>` — inside `<function>`

**Dependencies to add:**
  - `<package>@<version pinned in frontmatter>`

**Preserved behavior:** <list the existing repo behaviors that must
keep working after this edit — e.g., "existing OpenAI streaming still
works," "existing Redis session store still used," "existing tests
still pass unchanged.">

**Coexistence:** <one bullet per existing system the integration sits
alongside. Name the files/classes. Example: "The existing
`agents/memory/storage.py` MemoryStorage class remains untouched and
keeps its LangGraph SummarizationEvent flow. Mem0 is added as a
parallel long-term-facts store, in a new file, invoked only when
MEM0_ENABLED=1 is set.">

**Feature flag:** <the exact mechanism and the default. Required.
Example: `env MEM0_ENABLED=1`, default unset / off; `config.mem0.enabled`,
default false. With the flag in its default state, the repo must
behave exactly like `main`.>

**Sources consulted:** <minimum 2 URLs from "Canonical sources" above
that informed this plan. At least one `docs.mem0.ai` URL and one
delegated-skill URL. Cite the specific section or heading.>

**E2E recipe:** <how the verification skill should drive the app
end-to-end. Omit only if the repo is a pure library with no runnable
entry point — in which case the E2E step will skip with a warning.>

    start:               <shell command to launch the app locally,
                          using $PORT for any network port>
    ready_probe:         <one of: url=<URL> status=<code>  /
                          log="<substring to wait for>"  /
                          sleep=<seconds, last resort>>
    compose_services:    <optional: whitespace-separated service
                          names in docker-compose.yml to start first;
                          use label mem0-e2e: "true" to mark them>
    write_call:          <command that triggers the Mem0 write path
                          exactly once; ≤ 60s runtime>
    write_async_wait_ms: <milliseconds to wait after write_call for
                          async memory flush; default 0>
    read_call:           <command that triggers the Mem0 read path,
                          typically a fresh session / new request>
    read_assert:         <substring, regex, or jsonpath=<expr>=<value>
                          that MUST appear in read_call's output for
                          the E2E to pass. Derived from goal.md's
                          "What gets stored.">

**Rejected alternatives:** <briefly, 1–2 bullets — patterns the skill
considered but did not pick, and why. Helps the user decide.>

Rules:

• Show the user the proposed call sites with 10 lines of context around each before asking for approval.
• If the skill can't find a plausible call site for either write or read, it exits with code 5 and asks the user to name the file(s) manually (this is the "no fit here" signal — don't guess).
• Max 3 rejection rounds on the plan. On the 4th, exit code 5 with the last plan and the user's notes.
• If the user edits plan.md by hand, reload and re-confirm.

plan.md (not goal.md) is the contract the subagent implements against in step 8.

7. Tests first (TDD)

Main agent writes failing tests against goal.md in the repo's native test framework:

Test assertion shapes must match the canonical signatures:

• Platform method signatures: https://docs.mem0.ai/openapi.json (request body schemas for /v1/memories/ and /v1/memories/search/).
• OSS method signatures: the delegated skill named in plan.md (fetched from its raw URL) or skills/mem0/SKILL.md as the default.
• Do not hand-roll request shapes. If the delegated skill has an example block, lift it verbatim.

Minimum two test files (paths taken from plan.md call sites):

• test_mem0_write.<ext> — asserts add() is called at the Write call site with the right payload shape (Platform messages-array vs OSS string) and the right user_id source.
• test_mem0_read.<ext> — asserts search() runs before the Read call site and the result is wired into the LLM prompt / response path.

Tests MUST be importable with MEM0_API_KEY unset. This is the design pressure that forces step 8's lazy MemoryClient() / Memory() construction — eager module-level init hits the API on import and breaks pre-existing test collection when the key is missing.

Run the tests. They must fail. If they pass before any implementation, the tests are wrong — rewrite them.

8. Implementation (subagent, fresh context)

Spawn a subagent with:

• Inputs: the repo, goal.md, plan.md, the two test files, and direct URLs to: the delegated skill (from plan.md), the SDK source (pinned per mem0_tested_versions), https://docs.mem0.ai/llms.txt, and https://docs.mem0.ai/openapi.json.

• No access to main agent's reasoning trace or scratchpad.

• System prompt (verbatim):

  You are implementing a Mem0 integration for an existing repo.
  
  Read these first:
  - plan.md           (the mechanical contract)
  - goal.md           (the intent — do not change it)
  - the test files    (do not change them either)
  - <delegated skill raw URL from plan.md>
  - https://docs.mem0.ai/llms.txt
  - https://docs.mem0.ai/openapi.json  (Platform only)
  
  Constraints — all required, all enforced at review:
  
  1. Touch only the files named in plan.md's call sites, or add
     strictly new files.
  2. Do not remove or rename any existing symbol. Do not change
     any public signature.
  3. Do not modify any existing test.
  4. Gate every line of new Mem0 code behind the feature flag from
     plan.md. With the flag in its default state, the repo must
     behave exactly like `main` — byte-for-byte, including stdout
     and return values.
  5. Use only the <Platform | OSS> SDK surface. No new dependencies
     beyond those listed under plan.md's "Dependencies to add."
  6. Preserve everything listed under plan.md's "Preserved behavior"
     and "Coexistence."
  7. Lazy client construction. `MemoryClient()` validates the API
     key in `__init__` (it makes a network call). Never instantiate
     it at module-import time — construct on first use inside the
     request / handler path. The same rule applies to OSS `Memory()`,
     which can eagerly initialize embedding and LLM providers. Use
     a function-local singleton (`functools.lru_cache`, a module-level
     `_client = None` + getter, or DI scope) — never a top-level
     global. Eager init breaks the pre-existing test suite at
     collection time whenever the key is missing or invalid, which
     is a non-invasiveness violation.
  
  Implement the plan to make the new tests pass while all
  pre-existing tests continue to pass unchanged.
  

Subagent returns a diff. Main agent reviews against plan.md (the mechanical contract) and goal.md (the intent):

• Approved → apply the diff, commit.
• Rejected → return with specific, actionable feedback (not "try again").
• Max 3 review loops. Beyond that → exit code 4 with the last diff and reviewer feedback.

9. Commit + handoff

Create branch mem0-integrate/<short-goal-slug> and commit in separate commits so reviewers can cherry-pick:

1. mem0: add gated dependency — just the pyproject.toml / package.json change.
2. mem0: add integration module — the new file(s).
3. mem0: wire into <call site> — the call-site edit(s), still gated.
4. mem0: add tests — the new test files.

If --no-heal is set → print Run /mem0-test-integration to verify. and exit. Otherwise proceed to step 10.

10. Self-healing loop (default ON; disable with --no-heal)

Run /mem0-test-integration --ci in a subprocess. If scorecard.json reports overall: pass → done, exit 0.

Otherwise loop:

1. Categorize the failing check from scorecard.json. Route per category:

   • install / static_checks → dependency or import fix.
   • unit_tests → wiring or assertion fix.
   • smoke_test → API key or SDK call-shape fix.
   • e2e_test → recipe, flag-wiring, or integration-point fix.
   • Pre-existing test failure (test skill exit code 7, non_invasive: false in scorecard) → STOP. This is a non-invasiveness violation. Do NOT attempt to "fix" it (that breaks principle 3). Exit code 6 with rationale.

2. Spawn a remediation subagent, fresh context. Inputs: plan.md, goal.md, scorecard.md, scorecard.json, the last committed diff, and the relevant log file for the failing category (test-stdout.log / smoke-stdout.log / e2e-app.log / e2e-calls.log).

   System prompt (verbatim):

   You are fixing a failing Mem0 integration test.
   
   Non-negotiable constraints:
   - Do not modify test files.
   - Do not remove or rename any existing symbol or signature.
   - Do not change pre-existing behavior. The feature flag from
     plan.md must still default to OFF, and with the flag in its
     default state the repo must behave exactly like main.
   - Touch only the files named in plan.md's call sites, or add
     strictly new files.
   - Return the smallest possible diff that fixes the single
     failing check listed in scorecard.md. No drive-by cleanup.
   

3. Apply the diff; commit on the same branch with message mem0-heal: <category> attempt <N>. Do NOT amend earlier commits (reviewers need the heal trail).

4. Re-run /mem0-test-integration --ci. Outcomes:

   • overall: pass → done, exit 0.
   • Same check still failing → increment attempt counter; loop.
   • A different check now failing → regression. Revert the heal commit (git revert HEAD --no-edit), record the regression in .mem0-integration/heal-trace.md, exit code 6.

5. Bounded iterations. Default 3 attempts per failing category. Override with --heal-max N (hard cap 10). On exhaustion, exit 6 with the full attempt trace: each diff, each scorecard, final log tail.

6. Post-loop summary written to .mem0-integration/heal-trace.md: which category failed, how many attempts, each diff's intent, final status, and — on success — the delta from initial scorecard to final.

Artifacts (all under .mem0-integration/)

.mem0-integration/ is added to .gitignore on first run. Nothing is written outside this directory and the repo's source tree.

Modes

Invocation

/mem0-integrate                            # interactive, heal ON
/mem0-integrate --no-heal                  # stop after commit; manual verify
/mem0-integrate --heal-max 5               # cap heal attempts per category (default 3)
/mem0-integrate --product platform         # skip the product ask
/mem0-integrate --product oss
/mem0-integrate --ci                       # non-interactive (for test harness)

Exit codes

Explicitly out of scope

• Surveying the repo for fit points. Humans decide where Mem0 helps before invoking this skill.
• Replacing any existing memory / session / state system. Always additive and feature-flagged; see "Integration principles."
• Modifying pre-existing tests, even to "fix" them under self-heal. Tests that fail after integration with the flag unset are a non-invasiveness violation, not a bug to patch.
• Deciding Platform vs OSS silently. Always ask with a recommendation.
• Switching branches, pushing, or opening PRs. Commits locally and stops (or enters the heal loop, still local).
• Data migration between stores. Point user at migration/oss-to-platform docs if they ask.
• Provider selection beyond the default LLM for OSS. If they need a custom LLM / embedder / vector store, route to components/* docs and re-run step 4 with the new key.